Skip to content

P
Paddle

BaiXuePrincess / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
“84eb98dfbc6cfc10a245d459666b7fe81f013219”上不存在“...post_training_quantization/pytorch_yolo_series/eval.py”
提交 1f6e0448 编写于 作者: T typhoonzero
浏览文件
操作

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into prepare_pserver_executor

上级 44c29abd 49313d40
隐藏空白更改
内联 并排
Showing with 8285 addition and 556 deletion
benchmark/cluster/README.md
浏览文件 @ 1f6e0448
...@@ -36,11 +36,41 @@ ...@@ -36,11 +36,41 @@
- Trainer Count: 100 - Trainer Count: 100
- Metrics: mini-batch / sec - Metrics: mini-batch / sec
| Batch Size | 32 | 64 | 128 | 256 |
| -- | -- | -- | -- | -- | <table>
| PaddlePaddle Fluid | - | - | - | - | <thead>
| PaddlePaddle v2 | - | - | - | - | <tr>
| TensorFlow | - | - | - | - | <th>Batch Size </th>
<th> 32</th>
<th>64</th>
<th>128 </th>
<th>256</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>PaddlePaddle v2 </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>TensorFlow </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
</tbody>
</table>
### Measure the Performance for Different PServer Count ### Measure the Performance for Different PServer Count
...@@ -48,11 +78,41 @@ ...@@ -48,11 +78,41 @@
- Batch Size: 64 - Batch Size: 64
- Metrics: mini-batch / sec - Metrics: mini-batch / sec
| PServer Count | 10 | 20 | 40 | 60 |
| -- | -- | -- | -- | -- | <table>
| PaddlePaddle Fluid | - | - | - | - | <thead>
| PaddlePaddle v2 | - | - | - | - | <tr>
| TensorFlow | - | - | - | - | <th>PServer Count </th>
<th>10</th>
<th>20</th>
<th>40 </th>
<th>60</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>PaddlePaddle v2 </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>TensorFlow </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
</tr>
</tbody>
</table>
### Measure Parallel Efficiency By Increasing Trainer Count ### Measure Parallel Efficiency By Increasing Trainer Count
...@@ -67,11 +127,69 @@ The parallel efficiency is: ...@@ -67,11 +127,69 @@ The parallel efficiency is:
$E = \div(S, N)$ $E = \div(S, N)$
| Trainer Counter | 1 | 10 | 20 | 30 | 40 | 50 | 60 | 70 | 80 | 90 | 100 | <table>
| -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | <thead>
| PaddlePaddle Fluid | - | - | - | - | - | - | - | - | - | - | - | <tr>
| PaddlePaddle v2 | - | - | - | - | - | - | - | - | - | - | - | - | <th>Trainer Counter </th>
| TensorFlow | - | - | - | - | - | - | - | - | - | - | - | - | - | <th>1</th>
<th>10</th>
<th>20 </th>
<th>30</th>
<th>40</th>
<th>50</th>
<th>60 </th>
<th>70</th>
<th>80</th>
<th>90</th>
<th>100 </th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>PaddlePaddle v2 </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
</tr>
<tr>
<td>TensorFlow </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
<td>- </td>
<td>-</td>
<td>- </td>
<td>- </td>
</tr>
</tbody>
</table>
## Reproduce the benchmark ## Reproduce the benchmark
......
benchmark/cluster/vgg16/README.md
浏览文件 @ 1f6e0448
...@@ -16,11 +16,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`. ...@@ -16,11 +16,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Metrics: samples / sec - Metrics: samples / sec
| Batch Size | 32 | 64 | 128 | 256 | <table>
| -- | -- | -- | -- | -- | <thead>
| PaddlePaddle Fluid | 15.44 | 16.32 | 16.74 | 16.79 | <tr>
| PaddlePaddle v2 | 15.97 | 17.04 | 17.60 | 17.83 | <th>Batch Size </th>
| TensorFlow | 9.09 | 9.10 | 9.24 | 8.66 | <th> 32</th>
<th>64</th>
<th>128 </th>
<th>256</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td> 15.44 </td>
<td> 16.32 </td>
<td> 16.74 </td>
<td> 16.79 </td>
</tr>
<tr>
<td>PaddlePaddle v2 </td>
<td> 15.97 </td>
<td> 17.04 </td>
<td> 17.60 </td>
<td> 17.83 </td>
</tr>
<tr>
<td>TensorFlow </td>
<td> 9.09 </td>
<td> 9.10 </td>
<td> 9.24 </td>
<td> 8.66 </td>
</tr>
</tbody>
</table>
### Different Batch Size ### Different Batch Size
...@@ -28,12 +58,40 @@ Setting environment variable: `MKL_NUM_THREADS=1`. ...@@ -28,12 +58,40 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Trainer Count: 20 - Trainer Count: 20
- Metrics: samples / sec - Metrics: samples / sec
| Batch Size | 32 | 64 | 128 | 256 | <table>
| -- | -- | -- | -- | -- | <thead>
| PaddlePaddle Fluid | 190.20 | 222.15 | 247.40 | 258.18 | <tr>
| PaddlePaddle v2 | 170.96 | 233.71 | 256.14 | 329.23 | <th>Batch Size </th>
| TensorFlow | - | - | - | - | <th> 32</th>
<th>64</th>
<th>128 </th>
<th>256</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td> 190.20 </td>
<td> 222.15 </td>
<td> 247.40 </td>
<td> 258.18 </td>
</tr>
<tr>
<td>PaddlePaddle v2 </td>
<td> 170.96 </td>
<td> 233.71 </td>
<td> 256.14 </td>
<td> 329.23 </td>
</tr>
<tr>
<td>TensorFlow </td>
<td> - </td>
<td> - </td>
<td> - </td>
<td> - </td>
</tr>
</tbody>
</table>
### Accelerate Rate ### Accelerate Rate
...@@ -41,11 +99,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`. ...@@ -41,11 +99,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Batch Size: 128 - Batch Size: 128
- Metrics: samples / sec - Metrics: samples / sec
| Trainer Count | 20 | 40 | 80 | 100 | <table>
| -- | -- | -- | -- | -- | <thead>
| PaddlePaddle Fluid | 263.29 (78.64%) | 518.80 (77.47%) | 836.26 (62.44%) | 1019.29 (60.89%) | <tr>
| PaddlePaddle v2 (need more tests) | 326.85 (92.85%) | 534.58 (75.93%) | 853.30 (60.60%) | 1041.99 (59.20%) | <th>Trainer Count </th>
| TensorFlow | - | - | - | - | <th>20</th>
<th>40</th>
<th>80</th>
<th>100</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid</td>
<td> 263.29 (78.64%) </td>
<td> 518.80 (77.47%) </td>
<td> 836.26 (62.44%) </td>
<td> 1019.29 (60.89%) </td>
</tr>
<tr>
<td>PaddlePaddle v2 (need more tests) </td>
<td> 326.85 (92.85%) </td>
<td> 534.58 (75.93%) </td>
<td> 853.30 (60.60%) </td>
<td> 1041.99 (59.20%) </td>
</tr>
<tr>
<td>TensorFlow </td>
<td> - </td>
<td> - </td>
<td> - </td>
<td> - </td>
</tr>
</tbody>
</table>
### Different Pserver Count ### Different Pserver Count
...@@ -53,11 +141,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`. ...@@ -53,11 +141,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Batch Size: 128 - Batch Size: 128
- Metrics: samples/ sec - Metrics: samples/ sec
| PServer Count | 3 | 6 |10 | 20 | <table>
| -- | -- | -- | -- | -- | <thead>
| PaddlePaddle Fluid(should fix in next PR) | 589.1 | 592.6 | 656.4 | 655.8 | <tr>
| PaddlePaddle v2 | 593.4 | 791.3 | 729.7 | 821.7 | <th>PServer Count </th>
| TensorFlow | - | - | - | - | <th>3</th>
<th>6</th>
<th>10</th>
<th>20</th>
</tr>
</thead>
<tbody>
<tr>
<td> PaddlePaddle Fluid(should fix in next PR) </td>
<td> 589.1 </td>
<td> 592.6 </td>
<td> 656.4 </td>
<td> 655.8 </td>
</tr>
<tr>
<td>PaddlePaddle v2 (need more tests) </td>
<td> 593.4 </td>
<td> 791.3 </td>
<td> 729.7 </td>
<td> 821.7 </td>
</tr>
<tr>
<td>TensorFlow </td>
<td> - </td>
<td> - </td>
<td> - </td>
<td> - </td>
</tr>
</tbody>
</table>
*The performance gap between Fuild and v2 comes from the network interference.* *The performance gap between Fuild and v2 comes from the network interference.*
......
benchmark/fluid/machine_translation.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""seq2seq model for fluid."""
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import argparse
import time
import distutils.util
import paddle.v2 as paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import paddle.fluid.framework as framework
from paddle.fluid.executor import Executor
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument(
"--embedding_dim",
type=int,
default=512,
help="The dimension of embedding table. (default: %(default)d)")
parser.add_argument(
"--encoder_size",
type=int,
default=512,
help="The size of encoder bi-rnn unit. (default: %(default)d)")
parser.add_argument(
"--decoder_size",
type=int,
default=512,
help="The size of decoder rnn unit. (default: %(default)d)")
parser.add_argument(
"--batch_size",
type=int,
default=16,
help="The sequence number of a mini-batch data. (default: %(default)d)")
parser.add_argument(
"--dict_size",
type=int,
default=30000,
help="The dictionary capacity. Dictionaries of source sequence and "
"target dictionary have same capacity. (default: %(default)d)")
parser.add_argument(
"--pass_num",
type=int,
default=2,
help="The pass number to train. (default: %(default)d)")
parser.add_argument(
"--learning_rate",
type=float,
default=0.0002,
help="Learning rate used to train the model. (default: %(default)f)")
parser.add_argument(
"--infer_only", action='store_true', help="If set, run forward only.")
parser.add_argument(
"--beam_size",
type=int,
default=3,
help="The width for beam searching. (default: %(default)d)")
parser.add_argument(
"--use_gpu",
type=distutils.util.strtobool,
default=True,
help="Whether to use gpu. (default: %(default)d)")
parser.add_argument(
"--max_length",
type=int,
default=250,
help="The maximum length of sequence when doing generation. "
"(default: %(default)d)")
def lstm_step(x_t, hidden_t_prev, cell_t_prev, size):
def linear(inputs):
return fluid.layers.fc(input=inputs, size=size, bias_attr=True)
forget_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
input_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
output_gate = fluid.layers.sigmoid(x=linear([hidden_t_prev, x_t]))
cell_tilde = fluid.layers.tanh(x=linear([hidden_t_prev, x_t]))
cell_t = fluid.layers.sums(input=[
fluid.layers.elementwise_mul(
x=forget_gate, y=cell_t_prev), fluid.layers.elementwise_mul(
x=input_gate, y=cell_tilde)
])
hidden_t = fluid.layers.elementwise_mul(
x=output_gate, y=fluid.layers.tanh(x=cell_t))
return hidden_t, cell_t
def seq_to_seq_net(embedding_dim, encoder_size, decoder_size, source_dict_dim,
target_dict_dim, is_generating, beam_size, max_length):
"""Construct a seq2seq network."""
def bi_lstm_encoder(input_seq, gate_size):
# Linear transformation part for input gate, output gate, forget gate
# and cell activation vectors need be done outside of dynamic_lstm.
# So the output size is 4 times of gate_size.
input_forward_proj = fluid.layers.fc(input=input_seq,
size=gate_size * 4,
act=None,
bias_attr=False)
forward, _ = fluid.layers.dynamic_lstm(
input=input_forward_proj, size=gate_size * 4, use_peepholes=False)
input_reversed_proj = fluid.layers.fc(input=input_seq,
size=gate_size * 4,
act=None,
bias_attr=False)
reversed, _ = fluid.layers.dynamic_lstm(
input=input_reversed_proj,
size=gate_size * 4,
is_reverse=True,
use_peepholes=False)
return forward, reversed
src_word_idx = fluid.layers.data(
name='source_sequence', shape=[1], dtype='int64', lod_level=1)
src_embedding = fluid.layers.embedding(
input=src_word_idx,
size=[source_dict_dim, embedding_dim],
dtype='float32')
src_forward, src_reversed = bi_lstm_encoder(
input_seq=src_embedding, gate_size=encoder_size)
encoded_vector = fluid.layers.concat(
input=[src_forward, src_reversed], axis=1)
encoded_proj = fluid.layers.fc(input=encoded_vector,
size=decoder_size,
bias_attr=False)
backward_first = fluid.layers.sequence_pool(
input=src_reversed, pool_type='first')
decoder_boot = fluid.layers.fc(input=backward_first,
size=decoder_size,
bias_attr=False,
act='tanh')
def lstm_decoder_with_attention(target_embedding, encoder_vec, encoder_proj,
decoder_boot, decoder_size):
def simple_attention(encoder_vec, encoder_proj, decoder_state):
decoder_state_proj = fluid.layers.fc(input=decoder_state,
size=decoder_size,
bias_attr=False)
decoder_state_expand = fluid.layers.sequence_expand(
x=decoder_state_proj, y=encoder_proj)
concated = fluid.layers.concat(
input=[encoder_proj, decoder_state_expand], axis=1)
attention_weights = fluid.layers.fc(input=concated,
size=1,
act='tanh',
bias_attr=False)
attention_weights = fluid.layers.sequence_softmax(
input=attention_weights)
weigths_reshape = fluid.layers.reshape(
x=attention_weights, shape=[-1])
scaled = fluid.layers.elementwise_mul(
x=encoder_vec, y=weigths_reshape, axis=0)
context = fluid.layers.sequence_pool(input=scaled, pool_type='sum')
return context
rnn = fluid.layers.DynamicRNN()
cell_init = fluid.layers.fill_constant_batch_size_like(
input=decoder_boot,
value=0.0,
shape=[-1, decoder_size],
dtype='float32')
cell_init.stop_gradient = False
with rnn.block():
current_word = rnn.step_input(target_embedding)
encoder_vec = rnn.static_input(encoder_vec)
encoder_proj = rnn.static_input(encoder_proj)
hidden_mem = rnn.memory(init=decoder_boot, need_reorder=True)
cell_mem = rnn.memory(init=cell_init)
context = simple_attention(encoder_vec, encoder_proj, hidden_mem)
decoder_inputs = fluid.layers.concat(
input=[context, current_word], axis=1)
h, c = lstm_step(decoder_inputs, hidden_mem, cell_mem, decoder_size)
rnn.update_memory(hidden_mem, h)
rnn.update_memory(cell_mem, c)
out = fluid.layers.fc(input=h,
size=target_dict_dim,
bias_attr=True,
act='softmax')
rnn.output(out)
return rnn()
if not is_generating:
trg_word_idx = fluid.layers.data(
name='target_sequence', shape=[1], dtype='int64', lod_level=1)
trg_embedding = fluid.layers.embedding(
input=trg_word_idx,
size=[target_dict_dim, embedding_dim],
dtype='float32')
prediction = lstm_decoder_with_attention(trg_embedding, encoded_vector,
encoded_proj, decoder_boot,
decoder_size)
label = fluid.layers.data(
name='label_sequence', shape=[1], dtype='int64', lod_level=1)
cost = fluid.layers.cross_entropy(input=prediction, label=label)
avg_cost = fluid.layers.mean(x=cost)
feeding_list = ["source_sequence", "target_sequence", "label_sequence"]
return avg_cost, feeding_list
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = np.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
lod_t = core.LoDTensor()
lod_t.set(flattened_data, place)
lod_t.set_lod([lod])
return lod_t, lod[-1]
def lodtensor_to_ndarray(lod_tensor):
dims = lod_tensor.get_dims()
ndarray = np.zeros(shape=dims).astype('float32')
for i in xrange(np.product(dims)):
ndarray.ravel()[i] = lod_tensor.get_float_element(i)
return ndarray
def train():
avg_cost, feeding_list = seq_to_seq_net(
args.embedding_dim,
args.encoder_size,
args.decoder_size,
args.dict_size,
args.dict_size,
False,
beam_size=args.beam_size,
max_length=args.max_length)
# clone from default main program
inference_program = fluid.default_main_program().clone()
optimizer = fluid.optimizer.Adam(learning_rate=args.learning_rate)
optimizer.minimize(avg_cost)
fluid.memory_optimize(fluid.default_main_program())
train_batch_generator = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.wmt14.train(args.dict_size), buf_size=1000),
batch_size=args.batch_size)
test_batch_generator = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.wmt14.test(args.dict_size), buf_size=1000),
batch_size=args.batch_size)
place = core.CUDAPlace(0) if args.use_gpu else core.CPUPlace()
exe = Executor(place)
exe.run(framework.default_startup_program())
def do_validation():
total_loss = 0.0
count = 0
for batch_id, data in enumerate(test_batch_generator()):
src_seq = to_lodtensor(map(lambda x: x[0], data), place)[0]
trg_seq = to_lodtensor(map(lambda x: x[1], data), place)[0]
lbl_seq = to_lodtensor(map(lambda x: x[2], data), place)[0]
fetch_outs = exe.run(inference_program,
feed={
feeding_list[0]: src_seq,
feeding_list[1]: trg_seq,
feeding_list[2]: lbl_seq
},
fetch_list=[avg_cost],
return_numpy=False)
total_loss += lodtensor_to_ndarray(fetch_outs[0])[0]
count += 1
return total_loss / count
for pass_id in xrange(args.pass_num):
pass_start_time = time.time()
words_seen = 0
for batch_id, data in enumerate(train_batch_generator()):
src_seq, word_num = to_lodtensor(map(lambda x: x[0], data), place)
words_seen += word_num
trg_seq, word_num = to_lodtensor(map(lambda x: x[1], data), place)
words_seen += word_num
lbl_seq, _ = to_lodtensor(map(lambda x: x[2], data), place)
fetch_outs = exe.run(framework.default_main_program(),
feed={
feeding_list[0]: src_seq,
feeding_list[1]: trg_seq,
feeding_list[2]: lbl_seq
},
fetch_list=[avg_cost])
avg_cost_val = np.array(fetch_outs[0])
print('pass_id=%d, batch_id=%d, train_loss: %f' %
(pass_id, batch_id, avg_cost_val))
pass_end_time = time.time()
test_loss = do_validation()
time_consumed = pass_end_time - pass_start_time
words_per_sec = words_seen / time_consumed
print("pass_id=%d, test_loss: %f, words/s: %f, sec/pass: %f" %
(pass_id, test_loss, words_per_sec, time_consumed))
def infer():
pass
if __name__ == '__main__':
args = parser.parse_args()
if args.infer_only:
infer()
else:
train()
benchmark/fluid/mnist.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import numpy as np
import argparse
import time
import paddle.v2 as paddle
import paddle.fluid as fluid
import paddle.fluid.profiler as profiler
SEED = 1
DTYPE = "float32"
# random seed must set before configuring the network.
# fluid.default_startup_program().random_seed = SEED
def parse_args():
parser = argparse.ArgumentParser("mnist model benchmark.")
parser.add_argument(
'--batch_size', type=int, default=128, help='The minibatch size.')
parser.add_argument(
'--iterations', type=int, default=35, help='The number of minibatches.')
parser.add_argument(
'--pass_num', type=int, default=5, help='The number of passes.')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type.')
parser.add_argument(
'--infer_only', action='store_true', help='If set, run forward only.')
parser.add_argument(
'--use_cprof', action='store_true', help='If set, use cProfile.')
parser.add_argument(
'--use_nvprof',
action='store_true',
help='If set, use nvprof for CUDA.')
args = parser.parse_args()
return args
def print_arguments(args):
vars(args)['use_nvprof'] = (vars(args)['use_nvprof'] and
vars(args)['device'] == 'GPU')
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def cnn_model(data):
conv_pool_1 = fluid.nets.simple_img_conv_pool(
input=data,
filter_size=5,
num_filters=20,
pool_size=2,
pool_stride=2,
act="relu")
conv_pool_2 = fluid.nets.simple_img_conv_pool(
input=conv_pool_1,
filter_size=5,
num_filters=50,
pool_size=2,
pool_stride=2,
act="relu")
# TODO(dzhwinter) : refine the initializer and random seed settting
SIZE = 10
input_shape = conv_pool_2.shape
param_shape = [reduce(lambda a, b: a * b, input_shape[1:], 1)] + [SIZE]
scale = (2.0 / (param_shape[0]**2 * SIZE))**0.5
predict = fluid.layers.fc(
input=conv_pool_2,
size=SIZE,
act="softmax",
param_attr=fluid.param_attr.ParamAttr(
initializer=fluid.initializer.NormalInitializer(
loc=0.0, scale=scale)))
return predict
def eval_test(exe, batch_acc, batch_size_tensor, inference_program):
test_reader = paddle.batch(
paddle.dataset.mnist.test(), batch_size=args.batch_size)
test_pass_acc = fluid.average.WeightedAverage()
for batch_id, data in enumerate(test_reader()):
img_data = np.array(map(lambda x: x[0].reshape([1, 28, 28]),
data)).astype(DTYPE)
y_data = np.array(map(lambda x: x[1], data)).astype("int64")
y_data = y_data.reshape([len(y_data), 1])
acc, weight = exe.run(inference_program,
feed={"pixel": img_data,
"label": y_data},
fetch_list=[batch_acc, batch_size_tensor])
test_pass_acc.add(value=acc, weight=weight)
pass_acc = test_pass_acc.eval()
return pass_acc
def run_benchmark(model, args):
if args.use_cprof:
pr = cProfile.Profile()
pr.enable()
start_time = time.time()
# Input data
images = fluid.layers.data(name='pixel', shape=[1, 28, 28], dtype=DTYPE)
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
# Train program
predict = model(images)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
# Evaluator
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
# inference program
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
# Optimization
opt = fluid.optimizer.AdamOptimizer(
learning_rate=0.001, beta1=0.9, beta2=0.999)
opt.minimize(avg_cost)
fluid.memory_optimize(fluid.default_main_program())
# Initialize executor
place = fluid.CPUPlace() if args.device == 'CPU' else fluid.CUDAPlace(0)
exe = fluid.Executor(place)
# Parameter initialization
exe.run(fluid.default_startup_program())
# Reader
train_reader = paddle.batch(
paddle.dataset.mnist.train(), batch_size=args.batch_size)
accuracy = fluid.average.WeightedAverage()
for pass_id in range(args.pass_num):
accuracy.reset()
pass_start = time.time()
for batch_id, data in enumerate(train_reader()):
img_data = np.array(
map(lambda x: x[0].reshape([1, 28, 28]), data)).astype(DTYPE)
y_data = np.array(map(lambda x: x[1], data)).astype("int64")
y_data = y_data.reshape([len(y_data), 1])
start = time.time()
outs = exe.run(
fluid.default_main_program(),
feed={"pixel": img_data,
"label": y_data},
fetch_list=[avg_cost, batch_acc, batch_size_tensor]
) # The accuracy is the accumulation of batches, but not the current batch.
accuracy.add(value=outs[1], weight=outs[2])
end = time.time()
loss = np.array(outs[0])
acc = np.array(outs[1])
print("pass=%d, batch=%d, loss=%f, error=%f, elapse=%f" %
(pass_id, batch_id, loss, 1 - acc, (end - start) / 1000))
pass_end = time.time()
train_avg_acc = accuracy.eval()
test_avg_acc = eval_test(exe, batch_acc, batch_size_tensor,
inference_program)
print("pass=%d, train_avg_acc=%f, test_avg_acc=%f, elapse=%f" %
(pass_id, train_avg_acc, test_avg_acc,
(pass_end - pass_start) / 1000))
if __name__ == '__main__':
args = parse_args()
print_arguments(args)
if args.use_nvprof and args.device == 'GPU':
with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
run_benchmark(cnn_model, args)
else:
run_benchmark(cnn_model, args)
benchmark/fluid/resnet.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import functools
import numpy as np
import time
import cProfile, pstats, StringIO
import paddle.v2 as paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import paddle.fluid.profiler as profiler
def parse_args():
parser = argparse.ArgumentParser('Convolution model benchmark.')
parser.add_argument(
'--model',
type=str,
choices=['resnet_imagenet', 'resnet_cifar10'],
default='resnet_imagenet',
help='The model architecture.')
parser.add_argument(
'--batch_size', type=int, default=32, help='The minibatch size.')
parser.add_argument(
'--use_fake_data',
action='store_true',
help='use real data or fake data')
parser.add_argument(
'--skip_batch_num',
type=int,
default=5,
help='The first num of minibatch num to skip, for better performance test'
)
parser.add_argument(
'--iterations', type=int, default=80, help='The number of minibatches.')
parser.add_argument(
'--pass_num', type=int, default=100, help='The number of passes.')
parser.add_argument(
'--data_format',
type=str,
default='NCHW',
choices=['NCHW', 'NHWC'],
help='The data data_format, now only support NCHW.')
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help='The device type.')
parser.add_argument(
'--data_set',
type=str,
default='flowers',
choices=['cifar10', 'flowers'],
help='Optional dataset for benchmark.')
parser.add_argument(
'--infer_only', action='store_true', help='If set, run forward only.')
parser.add_argument(
'--use_cprof', action='store_true', help='If set, use cProfile.')
parser.add_argument(
'--use_nvprof',
action='store_true',
help='If set, use nvprof for CUDA.')
parser.add_argument(
'--with_test',
action='store_true',
help='If set, test the testset during training.')
args = parser.parse_args()
return args
def print_arguments(args):
vars(args)['use_nvprof'] = (vars(args)['use_nvprof'] and
vars(args)['device'] == 'GPU')
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
def conv_bn_layer(input, ch_out, filter_size, stride, padding, act='relu'):
conv1 = fluid.layers.conv2d(
input=input,
filter_size=filter_size,
num_filters=ch_out,
stride=stride,
padding=padding,
act=None,
bias_attr=False)
return fluid.layers.batch_norm(input=conv1, act=act)
def shortcut(input, ch_out, stride):
ch_in = input.shape[1] if args.data_format == 'NCHW' else input.shape[-1]
if ch_in != ch_out:
return conv_bn_layer(input, ch_out, 1, stride, 0, None)
else:
return input
def basicblock(input, ch_out, stride):
short = shortcut(input, ch_out, stride)
conv1 = conv_bn_layer(input, ch_out, 3, stride, 1)
conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1, act=None)
return fluid.layers.elementwise_add(x=short, y=conv2, act='relu')
def bottleneck(input, ch_out, stride):
short = shortcut(input, ch_out * 4, stride)
conv1 = conv_bn_layer(input, ch_out, 1, stride, 0)
conv2 = conv_bn_layer(conv1, ch_out, 3, 1, 1)
conv3 = conv_bn_layer(conv2, ch_out * 4, 1, 1, 0, act=None)
return fluid.layers.elementwise_add(x=short, y=conv3, act='relu')
def layer_warp(block_func, input, ch_out, count, stride):
res_out = block_func(input, ch_out, stride)
for i in range(1, count):
res_out = block_func(res_out, ch_out, 1)
return res_out
def resnet_imagenet(input, class_dim, depth=50, data_format='NCHW'):
cfg = {
18: ([2, 2, 2, 1], basicblock),
34: ([3, 4, 6, 3], basicblock),
50: ([3, 4, 6, 3], bottleneck),
101: ([3, 4, 23, 3], bottleneck),
152: ([3, 8, 36, 3], bottleneck)
}
stages, block_func = cfg[depth]
conv1 = conv_bn_layer(input, ch_out=64, filter_size=7, stride=2, padding=3)
pool1 = fluid.layers.pool2d(
input=conv1, pool_type='avg', pool_size=3, pool_stride=2)
res1 = layer_warp(block_func, pool1, 64, stages[0], 1)
res2 = layer_warp(block_func, res1, 128, stages[1], 2)
res3 = layer_warp(block_func, res2, 256, stages[2], 2)
res4 = layer_warp(block_func, res3, 512, stages[3], 2)
pool2 = fluid.layers.pool2d(
input=res4,
pool_size=7,
pool_type='avg',
pool_stride=1,
global_pooling=True)
out = fluid.layers.fc(input=pool2, size=class_dim, act='softmax')
return out
def resnet_cifar10(input, class_dim, depth=32, data_format='NCHW'):
assert (depth - 2) % 6 == 0
n = (depth - 2) // 6
conv1 = conv_bn_layer(
input=input, ch_out=16, filter_size=3, stride=1, padding=1)
res1 = layer_warp(basicblock, conv1, 16, n, 1)
res2 = layer_warp(basicblock, res1, 32, n, 2)
res3 = layer_warp(basicblock, res2, 64, n, 2)
pool = fluid.layers.pool2d(
input=res3, pool_size=8, pool_type='avg', pool_stride=1)
out = fluid.layers.fc(input=pool, size=class_dim, act='softmax')
return out
def run_benchmark(model, args):
if args.use_cprof:
pr = cProfile.Profile()
pr.enable()
if args.data_set == "cifar10":
class_dim = 10
if args.data_format == 'NCHW':
dshape = [3, 32, 32]
else:
dshape = [32, 32, 3]
else:
class_dim = 102
if args.data_format == 'NCHW':
dshape = [3, 224, 224]
else:
dshape = [224, 224, 3]
input = fluid.layers.data(name='data', shape=dshape, dtype='float32')
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
predict = model(input, class_dim)
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
optimizer = fluid.optimizer.Momentum(learning_rate=0.01, momentum=0.9)
opts = optimizer.minimize(avg_cost)
fluid.memory_optimize(fluid.default_main_program())
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.train(),
buf_size=5120),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.cifar.test10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.test(),
batch_size=args.batch_size)
def test(exe):
test_accuracy = fluid.average.WeightedAverage()
for batch_id, data in enumerate(test_reader()):
img_data = np.array(map(lambda x: x[0].reshape(dshape),
data)).astype("float32")
y_data = np.array(map(lambda x: x[1], data)).astype("int64")
y_data = y_data.reshape([-1, 1])
acc, weight = exe.run(inference_program,
feed={"data": img_data,
"label": y_data},
fetch_list=[batch_acc, batch_size_tensor])
test_accuracy.add(value=acc, weight=weight)
return test_accuracy.eval()
place = core.CPUPlace() if args.device == 'CPU' else core.CUDAPlace(0)
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
accuracy = fluid.average.WeightedAverage()
if args.use_fake_data:
data = train_reader().next()
image = np.array(map(lambda x: x[0].reshape(dshape), data)).astype(
'float32')
label = np.array(map(lambda x: x[1], data)).astype('int64')
label = label.reshape([-1, 1])
iters, num_samples, start_time = 0, 0, time.time()
for pass_id in range(args.pass_num):
accuracy.reset()
train_accs = []
train_losses = []
for batch_id, data in enumerate(train_reader()):
if iters == args.skip_batch_num:
start_time = time.time()
num_samples = 0
if iters == args.iterations:
break
if not args.use_fake_data:
image = np.array(map(lambda x: x[0].reshape(dshape),
data)).astype('float32')
label = np.array(map(lambda x: x[1], data)).astype('int64')
label = label.reshape([-1, 1])
loss, acc, weight = exe.run(
fluid.default_main_program(),
feed={'data': image,
'label': label},
fetch_list=[avg_cost, batch_acc, batch_size_tensor])
iters += 1
num_samples += label[0]
accuracy.add(value=acc, weight=weight)
train_losses.append(loss)
train_accs.append(acc)
print("Pass: %d, Iter: %d, Loss: %f, Accuracy: %f" %
(pass_id, iters, loss, acc))
pass_train_acc = accuracy.eval()
# evaluation
if args.with_test:
pass_test_acc = test(exe)
train_elapsed = time.time() - start_time
print("Pass: %d, Loss: %f, Train Accuray: %f\n" %
(pass_id, np.mean(train_losses), np.mean(train_accs)))
examples_per_sec = num_samples / train_elapsed
print('\nTotal examples: %d, total time: %.5f, %.5f examples/sed\n' %
(num_samples, train_elapsed, examples_per_sec))
if args.use_cprof:
pr.disable()
s = StringIO.StringIO()
sortby = 'cumulative'
ps = pstats.Stats(pr, stream=s).sort_stats(sortby)
ps.print_stats()
print(s.getvalue())
if __name__ == '__main__':
model_map = {
'resnet_imagenet': resnet_imagenet,
'resnet_cifar10': resnet_cifar10
}
args = parse_args()
print_arguments(args)
if args.data_format == 'NHWC':
raise ValueError('Only support NCHW data_format now.')
if args.use_nvprof and args.device == 'GPU':
with profiler.cuda_profiler("cuda_profiler.txt", 'csv') as nvprof:
run_benchmark(model_map[args.model], args)
else:
run_benchmark(model_map[args.model], args)
benchmark/fluid/run.sh 0 → 100644
浏览文件 @ 1f6e0448
#!/bin/bash
# This script benchmarking the PaddlePaddle Fluid on
# single thread single GPU.
export CUDNN_PATH=/paddle/cudnn_v5/cuda/lib
# disable openmp and mkl parallel
#https://github.com/PaddlePaddle/Paddle/issues/7199
export MKL_NUM_THREADS=1
export OMP_NUM_THREADS=1
ht=`lscpu |grep "per core"|awk -F':' '{print $2}'|xargs`
if [ $ht -eq 1 ]; then # HT is OFF
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,0,0"
fi
if [ -z "$OMP_DYNAMIC" ]; then
export OMP_DYNAMIC="FALSE"
fi
else # HT is ON
if [ -z "$KMP_AFFINITY" ]; then
export KMP_AFFINITY="granularity=fine,compact,1,0"
fi
fi
# disable multi-gpu if have more than one
export CUDA_VISIBLE_DEVICES=0
export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=$CUDNN_PATH:$LD_LIBRARY_PATH
# vgg16
# cifar10 gpu cifar10 128
FLAGS_benchmark=true python fluid/vgg.py \
--device=GPU \
--batch_size=128 \
--skip_batch_num=5 \
--iterations=30 \
2>&1 > vgg16_gpu_128.log
# resnet50
# resnet50 gpu cifar10 128
FLAGS_benchmark=true python fluid/resnet.py \
--device=GPU \
--batch_size=128 \
--data_set=cifar10 \
--model=resnet_cifar10 \
--skip_batch_num=5 \
--iterations=30 \
2>&1 > resnet50_gpu_128.log
# lstm
benchmark/fluid/stacked_dynamic_lstm.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
import argparse
import cPickle
import os
import random
import time
import numpy
import paddle.v2 as paddle
import paddle.v2.dataset.imdb as imdb
import paddle.fluid as fluid
from paddle.v2 import batch
import paddle.fluid.profiler as profiler
def parse_args():
parser = argparse.ArgumentParser("Understand Sentiment by Dynamic RNN.")
parser.add_argument(
'--batch_size',
type=int,
default=32,
help='The sequence number of a batch data. (default: %(default)d)')
parser.add_argument(
'--emb_dim',
type=int,
default=512,
help='Dimension of embedding table. (default: %(default)d)')
parser.add_argument(
'--hidden_dim',
type=int,
default=512,
help='Hidden size of lstm unit. (default: %(default)d)')
parser.add_argument(
'--pass_num',
type=int,
default=100,
help='Epoch number to train. (default: %(default)d)')
parser.add_argument(
'--device',
type=str,
default='CPU',
choices=['CPU', 'GPU'],
help='The device type.')
parser.add_argument(
'--crop_size',
type=int,
default=int(os.environ.get('CROP_SIZE', '1500')),
help='The max sentence length of input. Since this model use plain RNN,'
' Gradient could be explored if sentence is too long')
args = parser.parse_args()
return args
word_dict = imdb.word_dict()
def crop_sentence(reader, crop_size):
unk_value = word_dict['<unk>']
def __impl__():
for item in reader():
if len([x for x in item[0] if x != unk_value]) < crop_size:
yield item
return __impl__
def main():
args = parse_args()
lstm_size = args.hidden_dim
data = fluid.layers.data(
name="words", shape=[1], lod_level=1, dtype='int64')
sentence = fluid.layers.embedding(
input=data, size=[len(word_dict), args.emb_dim])
sentence = fluid.layers.fc(input=sentence, size=lstm_size, act='tanh')
rnn = fluid.layers.DynamicRNN()
with rnn.block():
word = rnn.step_input(sentence)
prev_hidden = rnn.memory(value=0.0, shape=[lstm_size])
prev_cell = rnn.memory(value=0.0, shape=[lstm_size])
def gate_common(
ipt,
hidden,
size, ):
gate0 = fluid.layers.fc(input=ipt, size=size, bias_attr=True)
gate1 = fluid.layers.fc(input=hidden, size=size, bias_attr=False)
gate = fluid.layers.sums(input=[gate0, gate1])
return gate
forget_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
input_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
output_gate = fluid.layers.sigmoid(
x=gate_common(word, prev_hidden, lstm_size))
cell_gate = fluid.layers.tanh(
x=gate_common(word, prev_hidden, lstm_size))
cell = fluid.layers.sums(input=[
fluid.layers.elementwise_mul(
x=forget_gate, y=prev_cell), fluid.layers.elementwise_mul(
x=input_gate, y=cell_gate)
])
hidden = fluid.layers.elementwise_mul(
x=output_gate, y=fluid.layers.tanh(x=cell))
rnn.update_memory(prev_cell, cell)
rnn.update_memory(prev_hidden, hidden)
rnn.output(hidden)
last = fluid.layers.sequence_pool(rnn(), 'last')
logit = fluid.layers.fc(input=last, size=2, act='softmax')
loss = fluid.layers.cross_entropy(
input=logit,
label=fluid.layers.data(
name='label', shape=[1], dtype='int64'))
loss = fluid.layers.mean(x=loss)
# add acc
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(input=logit, label=fluid.layers.data(name='label', \
shape=[1], dtype='int64'), total=batch_size_tensor)
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
adam = fluid.optimizer.Adam()
adam.minimize(loss)
fluid.memory_optimize(fluid.default_main_program())
place = fluid.CPUPlace() if args.device == 'CPU' else fluid.CUDAPlace(0)
exe = fluid.Executor(place)
exe.run(fluid.default_startup_program())
def train_loop(pass_num, crop_size):
with profiler.profiler(args.device, 'total') as prof:
for pass_id in range(pass_num):
train_reader = batch(
paddle.reader.shuffle(
crop_sentence(imdb.train(word_dict), crop_size),
buf_size=25000),
batch_size=args.batch_size)
word_nums = 0
pass_start_time = time.time()
for batch_id, data in enumerate(train_reader()):
tensor_words = to_lodtensor([x[0] for x in data], place)
for x in data:
word_nums += len(x[0])
label = numpy.array([x[1] for x in data]).astype("int64")
label = label.reshape((-1, 1))
loss_np, acc, weight = exe.run(
fluid.default_main_program(),
feed={"words": tensor_words,
"label": label},
fetch_list=[loss, batch_acc, batch_size_tensor])
print("pass_id=%d, batch_id=%d, loss=%f, acc=%f" %
(pass_id, batch_id, loss_np, acc))
pass_end_time = time.time()
time_consumed = pass_end_time - pass_start_time
words_per_sec = word_nums / time_consumed
print("pass_id=%d, sec/pass: %f, words/s: %f" %
(pass_id, time_consumed, words_per_sec))
train_loop(args.pass_num, args.crop_size)
def to_lodtensor(data, place):
seq_lens = [len(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
cur_len += l
lod.append(cur_len)
flattened_data = numpy.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
res = fluid.LoDTensor()
res.set(flattened_data, place)
res.set_lod([lod])
return res
if __name__ == '__main__':
main()
benchmark/fluid/vgg.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""VGG16 benchmark in Fluid"""
from __future__ import print_function
import sys
import time
import numpy as np
import paddle.v2 as paddle
import paddle.fluid as fluid
import paddle.fluid.core as core
import argparse
import functools
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument(
'--batch_size', type=int, default=128, help="Batch size for training.")
parser.add_argument(
'--skip_batch_num',
type=int,
default=5,
help='The first num of minibatch num to skip, for better performance test')
parser.add_argument(
'--iterations', type=int, default=80, help='The number of minibatches.')
parser.add_argument(
'--learning_rate',
type=float,
default=1e-3,
help="Learning rate for training.")
parser.add_argument('--pass_num', type=int, default=50, help="No. of passes.")
parser.add_argument(
'--device',
type=str,
default='GPU',
choices=['CPU', 'GPU'],
help="The device type.")
parser.add_argument(
'--data_format',
type=str,
default='NCHW',
choices=['NCHW', 'NHWC'],
help='The data order, now only support NCHW.')
parser.add_argument(
'--data_set',
type=str,
default='cifar10',
choices=['cifar10', 'flowers'],
help='Optional dataset for benchmark.')
parser.add_argument(
'--with_test',
action='store_true',
help='If set, test the testset during training.')
args = parser.parse_args()
def vgg16_bn_drop(input):
def conv_block(input, num_filter, groups, dropouts):
return fluid.nets.img_conv_group(
input=input,
pool_size=2,
pool_stride=2,
conv_num_filter=[num_filter] * groups,
conv_filter_size=3,
conv_act='relu',
conv_with_batchnorm=True,
conv_batchnorm_drop_rate=dropouts,
pool_type='max')
conv1 = conv_block(input, 64, 2, [0.3, 0])
conv2 = conv_block(conv1, 128, 2, [0.4, 0])
conv3 = conv_block(conv2, 256, 3, [0.4, 0.4, 0])
conv4 = conv_block(conv3, 512, 3, [0.4, 0.4, 0])
conv5 = conv_block(conv4, 512, 3, [0.4, 0.4, 0])
drop = fluid.layers.dropout(x=conv5, dropout_prob=0.5)
fc1 = fluid.layers.fc(input=drop, size=512, act=None)
bn = fluid.layers.batch_norm(input=fc1, act='relu')
drop2 = fluid.layers.dropout(x=bn, dropout_prob=0.5)
fc2 = fluid.layers.fc(input=drop2, size=512, act=None)
return fc2
def main():
if args.data_set == "cifar10":
classdim = 10
if args.data_format == 'NCHW':
data_shape = [3, 32, 32]
else:
data_shape = [32, 32, 3]
else:
classdim = 102
if args.data_format == 'NCHW':
data_shape = [3, 224, 224]
else:
data_shape = [224, 224, 3]
# Input data
images = fluid.layers.data(name='pixel', shape=data_shape, dtype='float32')
label = fluid.layers.data(name='label', shape=[1], dtype='int64')
# Train program
net = vgg16_bn_drop(images)
predict = fluid.layers.fc(input=net, size=classdim, act='softmax')
cost = fluid.layers.cross_entropy(input=predict, label=label)
avg_cost = fluid.layers.mean(x=cost)
# Evaluator
batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
batch_acc = fluid.layers.accuracy(
input=predict, label=label, total=batch_size_tensor)
# inference program
inference_program = fluid.default_main_program().clone()
with fluid.program_guard(inference_program):
inference_program = fluid.io.get_inference_program(
target_vars=[batch_acc, batch_size_tensor])
# Optimization
optimizer = fluid.optimizer.Adam(learning_rate=args.learning_rate)
opts = optimizer.minimize(avg_cost)
fluid.memory_optimize(fluid.default_main_program())
# Initialize executor
place = core.CPUPlace() if args.device == 'CPU' else core.CUDAPlace(0)
exe = fluid.Executor(place)
# Parameter initialization
exe.run(fluid.default_startup_program())
# data reader
train_reader = paddle.batch(
paddle.reader.shuffle(
paddle.dataset.cifar.train10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.train(),
buf_size=5120),
batch_size=args.batch_size)
test_reader = paddle.batch(
paddle.dataset.cifar.test10()
if args.data_set == 'cifar10' else paddle.dataset.flowers.test(),
batch_size=args.batch_size)
# test
def test(exe):
test_accuracy = fluid.average.WeightedAverage()
for batch_id, data in enumerate(test_reader()):
img_data = np.array(map(lambda x: x[0].reshape(data_shape),
data)).astype("float32")
y_data = np.array(map(lambda x: x[1], data)).astype("int64")
y_data = y_data.reshape([-1, 1])
acc, weight = exe.run(inference_program,
feed={"pixel": img_data,
"label": y_data},
fetch_list=[batch_acc, batch_size_tensor])
test_accuracy.add(value=acc, weight=weight)
return test_accuracy.eval()
iters, num_samples, start_time = 0, 0, time.time()
accuracy = fluid.average.WeightedAverage()
for pass_id in range(args.pass_num):
accuracy.reset()
train_accs = []
train_losses = []
for batch_id, data in enumerate(train_reader()):
if iters == args.skip_batch_num:
start_time = time.time()
num_samples = 0
if iters == args.iterations:
break
img_data = np.array(map(lambda x: x[0].reshape(data_shape),
data)).astype("float32")
y_data = np.array(map(lambda x: x[1], data)).astype("int64")
y_data = y_data.reshape([-1, 1])
loss, acc, weight = exe.run(
fluid.default_main_program(),
feed={"pixel": img_data,
"label": y_data},
fetch_list=[avg_cost, batch_acc, batch_size_tensor])
accuracy.add(value=acc, weight=weight)
iters += 1
num_samples += len(data)
print(
"Pass = %d, Iter = %d, Loss = %f, Accuracy = %f" %
(pass_id, iters, loss, acc)
) # The accuracy is the accumulation of batches, but not the current batch.
pass_train_acc = accuracy.eval()
train_losses.append(loss)
train_accs.append(acc)
# evaluation
if args.with_test:
pass_test_acc = test(exe)
train_elapsed = time.time() - start_time
print("Pass: %d, Loss: %f, Train Accuray: %f\n" %
(pass_id, np.mean(train_losses), np.mean(train_accs)))
def print_arguments():
print('----------- Configuration Arguments -----------')
for arg, value in sorted(vars(args).iteritems()):
print('%s: %s' % (arg, value))
print('------------------------------------------------')
if __name__ == "__main__":
print_arguments()
main()
doc/fluid/api/layers.rst
浏览文件 @ 1f6e0448
...@@ -494,6 +494,12 @@ reshape ...@@ -494,6 +494,12 @@ reshape
.. autofunction:: paddle.fluid.layers.reshape .. autofunction:: paddle.fluid.layers.reshape
:noindex: :noindex:
pad
---
.. autofunction:: paddle.fluid.layers.pad
:noindex:
scale scale
----- -----
......
doc/fluid/design/algorithm/parameter_average.md
浏览文件 @ 1f6e0448
...@@ -7,7 +7,7 @@ Polyak and Juditsky (1992) showed that the test performance of simple average of ...@@ -7,7 +7,7 @@ Polyak and Juditsky (1992) showed that the test performance of simple average of
Hence, to accelerate the speed of Stochastic Gradient Descent, Averaged Stochastic Gradient Descent (ASGD) was proposed in Polyak and Juditsky (1992). For ASGD, the running average of parameters obtained by SGD, is used as the estimator for <img src="./images/theta_star.gif"/><br/> . The averaging is done as follows: Hence, to accelerate the speed of Stochastic Gradient Descent, Averaged Stochastic Gradient Descent (ASGD) was proposed in Polyak and Juditsky (1992). For ASGD, the running average of parameters obtained by SGD, is used as the estimator for <img src="./images/theta_star.gif"/><br/> . The averaging is done as follows:
<img src="./images/asgd.gif" align="center"/><br/> ![](./images/asgd.gif)
We propose averaging for any optimizer similar to how ASGD performs it, as mentioned above. We propose averaging for any optimizer similar to how ASGD performs it, as mentioned above.
......
doc/fluid/design/concepts/README.md
浏览文件 @ 1f6e0448
...@@ -6,11 +6,33 @@ Here are some initial thoughts. Your comments are welcome! ...@@ -6,11 +6,33 @@ Here are some initial thoughts. Your comments are welcome!
I think we need only the following few CMake functions to make a project description mean and clean: I think we need only the following few CMake functions to make a project description mean and clean:
| C++ | CUDA C++ | Go | <table>
|---|---|---| <thead>
| cc_library | nv_library | go_library | <tr>
| cc_binary | nv_binary | go_binary | <th>C++</th>
| cc_test | nv_test | go_test | <th>CUDA C++</th>
<th>Go</th>
</tr>
</thead>
<tbody>
<tr>
<td>cc_library </td>
<td>nv_library </td>
<td>go_library </td>
</tr>
<tr>
<td>cc_binary </td>
<td>nv_binary </td>
<td>go_binary </td>
</tr>
<tr>
<td> cc_test </td>
<td> nv_test </td>
<td> go_test </td>
</tr>
</tbody>
</table>
- The `_library` functions generate .a files from source code. - The `_library` functions generate .a files from source code.
- The `_binary` functions generate executable binary files. - The `_binary` functions generate executable binary files.
......
doc/fluid/design/concepts/block.md
浏览文件 @ 1f6e0448
...@@ -14,11 +14,29 @@ In programming languages, a block is a pair of curly braces that includes local ...@@ -14,11 +14,29 @@ In programming languages, a block is a pair of curly braces that includes local
Blocks work with control flow structures like `if`, `else`, and `for`, which have equivalents in deep learning: Blocks work with control flow structures like `if`, `else`, and `for`, which have equivalents in deep learning:
| programming languages | PaddlePaddle | <table>
|-----------------------|-----------------------| <thead>
| for, while loop | RNN, WhileOp | <tr>
| if, if-else, switch | IfElseOp, SwitchOp | <th>programming languages</th>
| sequential execution | a sequence of layers | <th>PaddlePaddle</th>
</tr>
</thead>
<tbody>
<tr>
<td>for, while loop </td>
<td>RNN, WhileOp </td>
</tr>
<tr>
<td>if, if-else, switch </td>
<td>IfElseOp, SwitchOp </td>
</tr>
<tr>
<td>sequential execution </td>
<td>a sequence of layers </td>
</tr>
</tbody>
</table>
A key difference is that a C++ program describes a one pass computation, whereas a deep learning program describes both the forward and backward passes. A key difference is that a C++ program describes a one pass computation, whereas a deep learning program describes both the forward and backward passes.
...@@ -26,12 +44,33 @@ A key difference is that a C++ program describes a one pass computation, whereas ...@@ -26,12 +44,33 @@ A key difference is that a C++ program describes a one pass computation, whereas
The existence of the backward pass makes the execution of a block of PaddlePaddle different from traditional programs: The existence of the backward pass makes the execution of a block of PaddlePaddle different from traditional programs:
| programming languages | PaddlePaddle | <table>
|-----------------------|---------------------------------| <thead>
| stack | scope hierarchy | <tr>
| stack frame | scope | <th>programming languages</th>
| push at entering block| push at entering block | <th>PaddlePaddle</th>
| pop at leaving block | destroy when minibatch completes| </tr>
</thead>
<tbody>
<tr>
<td>stack </td>
<td>scope hierarchy </td>
</tr>
<tr>
<td>stack frame </td>
<td>scope </td>
</tr>
<tr>
<td>push at entering block </td>
<td>push at entering block </td>
</tr>
<tr>
<td>pop at leaving block </td>
<td>destroy when minibatch completes </td>
</tr>
</tbody>
</table>
1. In traditional programs: 1. In traditional programs:
......
doc/fluid/design/concepts/functions_operators_layers.md
浏览文件 @ 1f6e0448
...@@ -86,12 +86,40 @@ def layer.fc(X): ...@@ -86,12 +86,40 @@ def layer.fc(X):
We'd like to have Python bindings to operators in package `paddle.operator`, and Python compositions of operators in package `paddle.layer`. So we have the following concepts in above illustrative example: We'd like to have Python bindings to operators in package `paddle.operator`, and Python compositions of operators in package `paddle.layer`. So we have the following concepts in above illustrative example:
<table>
| C++ functions/functors | mul | add | | | <thead>
|------------------------|--------------|--------------|-------------|----------| <tr>
| C++ operator class | mulOp | addOp | FCOp | | <th>C++ functions/functors</th>
| Python binding | operator.mul | operator.add | operator.fc | | <th>mul</th>
| Python function | | | | layer.fc | <th>add</th>
<th></th>
<th></th>
</tr>
</thead>
<tbody>
<tr>
<td>C++ operator class </td>
<td>mulOp</td>
<td>addOp </td>
<td>FCOp </td>
<td></td>
</tr>
<tr>
<td>Python binding </td>
<td>operator.mul</td>
<td> operator.add </td>
<td>operator.fc </td>
<td></td>
</tr>
<tr>
<td>Python function </td>
<td></td>
<td></td>
<td> </td>
<td>layer.fc</td>
</tr>
</tbody>
</table>
This is how we differentiate layer and operators in PaddlePaddle: This is how we differentiate layer and operators in PaddlePaddle:
......
doc/fluid/design/concepts/lod_tensor.md
浏览文件 @ 1f6e0448
...@@ -2,12 +2,38 @@ ...@@ -2,12 +2,38 @@
Like other deep learning systems, PaddlePaddle supports training models from sequence data. Also, like other systems, PaddlePaddle represent a mini-batch of sequences as a Tensor. What is different is that PaddlePaddle doesn't require all sequences in a mini-batch to be of the same length. Thus no need for padding zeros. Like other deep learning systems, PaddlePaddle supports training models from sequence data. Also, like other systems, PaddlePaddle represent a mini-batch of sequences as a Tensor. What is different is that PaddlePaddle doesn't require all sequences in a mini-batch to be of the same length. Thus no need for padding zeros.
| | TensorFlow | PaddlePaddle | <table>
|-----------------------|------------|--------------| <thead>
| RNN | Support | Support | <tr>
| recursive RNN | Support | Support | <th></th>
| padding zeros | Must | No need | <th>TensorFlow</th>
| blob data type | Tensor | LoDTensor | <th>PaddlePaddle</th>
</tr>
</thead>
<tbody>
<tr>
<td>RNN </td>
<td>Support </td>
<td>Support </td>
</tr>
<tr>
<td>recursive RNN </td>
<td>Support </td>
<td>Support </td>
</tr>
<tr>
<td>padding zeros </td>
<td> Must </td>
<td>No need </td>
</tr>
<tr>
<td> blob data type </td>
<td> Tensor</td>
<td> LoDTensor </td>
</tr>
</tbody>
</table>
PaddlePaddle achieves this flexibility by passing through a new data type, *LoD Tensor*, which is a Tensor attached with segmentation index known as *LoD*, between operators. The LoD index doesn't only segment a tensor, but also recursively segments sub-sequences. This document presents the design of LoD and LoDTensor. PaddlePaddle achieves this flexibility by passing through a new data type, *LoD Tensor*, which is a Tensor attached with segmentation index known as *LoD*, between operators. The LoD index doesn't only segment a tensor, but also recursively segments sub-sequences. This document presents the design of LoD and LoDTensor.
......
doc/fluid/design/concepts/var_desc.md
浏览文件 @ 1f6e0448
...@@ -10,10 +10,27 @@ PaddlePaddle uses proto message to describe compile time program because : ...@@ -10,10 +10,27 @@ PaddlePaddle uses proto message to describe compile time program because :
The computation `Program` consists of nested `Blocks`. Each `Block` will consist of data(i.e. `Variable`) and `Operations`. The concept to represent them is in the table below. The computation `Program` consists of nested `Blocks`. Each `Block` will consist of data(i.e. `Variable`) and `Operations`. The concept to represent them is in the table below.
| |compile time|runtime| <table>
|---|---|---| <thead>
|Data|VarDesc(proto)|Variable(cpp)| <tr>
|Operation|OpDesc(proto)|Operator(cpp)| <th></th>
<th>compile time</th>
<th>runtime</th>
</tr>
</thead>
<tbody>
<tr>
<td>Data </td>
<td>VarDesc(proto) </td>
<td>Variable(cpp) </td>
</tr>
<tr>
<td>Operation </td>
<td>OpDesc(proto) </td>
<td>Operator(cpp) </td>
</tr>
</tbody>
</table>
## Definition of VarType ## Definition of VarType
......
doc/fluid/design/concurrent/concurrent_programming.md
浏览文件 @ 1f6e0448
...@@ -10,12 +10,38 @@ The answer relies on the fact that a `ProgramDesc` is similar to an abstract syn ...@@ -10,12 +10,38 @@ The answer relies on the fact that a `ProgramDesc` is similar to an abstract syn
The following table compares concepts in Fluid and Go The following table compares concepts in Fluid and Go
| Go | Fluid | <table>
|----|-------| <thead>
|user-defined functions | [layers](https://github.com/PaddlePaddle/Paddle/tree/develop/python/paddle/fluid) | <tr>
| control-flow and built-in functions | [intrinsics/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators) | <th></th>
| goroutines, channels | [class ThreadPool](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework/thread_pool.h) | <th>Go</th>
| runtime | [class Executor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/executor.h) | <th>Fluid</th>
</tr>
</thead>
<tbody>
<tr>
<td>user-defined functions </td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/tree/develop/python/paddle/fluid">layers</a></td>
</tr>
<tr>
<td>control-flow and built-in functions </td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators">intrinsics/operators</a></td>
</tr>
<tr>
<td>goroutines, channels </td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework/thread_pool.h">class ThreadPool</a></td>
</tr>
<tr>
<td>runtime </td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/executor.h">class Executor</a></td>
</tr>
</tbody>
</table>
## An Example Concurrent Program ## An Example Concurrent Program
...@@ -77,11 +103,11 @@ message ProgramDesc { ...@@ -77,11 +103,11 @@ message ProgramDesc {
read(output = X) read(output = X)
kube_get_workers_addrs(output = L) kube_get_workers_addrs(output = L)
Y = tensor_array(len(L)) Y = tensor_array(len(L))
parallel_for(input = X, output = Y, parallel_for(input = X, output = Y,
attrs = {L, block_id(1)}) # referring to block 1 attrs = {L, block_id(1)}) # referring to block 1
] ]
} }
block[1] = Block { block[1] = Block {
parent = 0, parent = 0,
vars = [x, y, index], vars = [x, y, index],
...@@ -102,7 +128,7 @@ func main() { //// block 0 ...@@ -102,7 +128,7 @@ func main() { //// block 0
X = fluid.read(...) X = fluid.read(...)
L = fluid.k8s.get_worker_addrs() L = fluid.k8s.get_worker_addrs()
Y = fluid.tensor_array(len(L)) Y = fluid.tensor_array(len(L))
fluid.parallel_for(X, L, fluid.parallel_for(X, L,
func(index int) { //// block 1 func(index int) { //// block 1
x = X[index] x = X[index]
fluid.send(L[index], x) fluid.send(L[index], x)
...@@ -116,7 +142,7 @@ An explanation of the above program: ...@@ -116,7 +142,7 @@ An explanation of the above program:
- `fluid.k8s` is a package that provides access to Kubernetes API. - `fluid.k8s` is a package that provides access to Kubernetes API.
- `fluid.k8s.get_worker_addrs` returns the list of IP and ports of all pods of the current job except for the current one (the master pod). - `fluid.k8s.get_worker_addrs` returns the list of IP and ports of all pods of the current job except for the current one (the master pod).
- `fluid.tensor_array` creates a [tensor array](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor_array.h). `fluid.parallel_for` creates a `ParallelFor` intrinsic, which, when executed, - `fluid.tensor_array` creates a [tensor array](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor_array.h). `fluid.parallel_for` creates a `ParallelFor` intrinsic, which, when executed,
1. creates `len(L)` scopes, each for the concurrent running of the sub-block (block 1 in this case), and initializes a variable named "index" in the scope to an integer value in the range `[0, len(L)-1]`, and 1. creates `len(L)` scopes, each for the concurrent running of the sub-block (block 1 in this case), and initializes a variable named "index" in the scope to an integer value in the range `[0, len(L)-1]`, and
2. creates `len(L)` threads by calling into the `ThreadPool` singleton, each thread 2. creates `len(L)` threads by calling into the `ThreadPool` singleton, each thread
......
doc/fluid/design/concurrent/csp.md
浏览文件 @ 1f6e0448
...@@ -13,14 +13,41 @@ Most DL systems, including TensorFlow, Caffe2, and MxNet, can asynchronously exe ...@@ -13,14 +13,41 @@ Most DL systems, including TensorFlow, Caffe2, and MxNet, can asynchronously exe
There were many concurrent programming models, implemented in various forms: There were many concurrent programming models, implemented in various forms:
| concurrent programming model | implementation | <table>
|-----|-----| <thead>
| mutex | types and functions in standard libraries | <tr>
| semaphore | types and functions in standard libraries | <th>concurrent programming model</th>
| communicating sequential processes (CSP) | Go programming language | <th>implementation</th>
| actor model | Erlang programming language | </tr>
| message passing | MPI | </thead>
| bulk synchronous parallel (BSP) | Pregel distributed programming framework | <tbody>
<tr>
<td>mutex </td>
<td>types and functions in standard libraries </td>
</tr>
<tr>
<td>semaphore </td>
<td> types and functions in standard libraries </td>
</tr>
<tr>
<td> communicating sequential processes (CSP) </td>
<td> Go programming language </td>
</tr>
<tr>
<td> actor model </td>
<td> Erlang programming language </td>
</tr>
<tr>
<td> message passing </td>
<td> MPI </td>
</tr>
<tr>
<td> bulk synchronous parallel (BSP) </td>
<td> Pregel distributed programming framework </td>
</tr>
</tbody>
</table>
Since Fluid was designed to be a programming language, we would like to implement CSP in Fluid. Since Fluid was designed to be a programming language, we would like to implement CSP in Fluid.
...@@ -118,9 +145,9 @@ There are four types of actions with a channel: ...@@ -118,9 +145,9 @@ There are four types of actions with a channel:
```go ```go
close(ch) close(ch)
``` ```
Please be aware that a closed channel is not a nil channel, which is `var ch chan int`. Please be aware that a closed channel is not a nil channel, which is `var ch chan int`.
There are some [axioms with channels](https://dave.cheney.net/2014/03/19/channel-axioms): There are some [axioms with channels](https://dave.cheney.net/2014/03/19/channel-axioms):
1. A send to a nil channel blocks forever 1. A send to a nil channel blocks forever
......
doc/fluid/design/modules/python_api.md
浏览文件 @ 1f6e0448
...@@ -2,12 +2,33 @@ ...@@ -2,12 +2,33 @@
Due to the refactorization of the PaddlePaddle core, we need Python classes to construct corresponding protobuf messages that describe a DL program. Due to the refactorization of the PaddlePaddle core, we need Python classes to construct corresponding protobuf messages that describe a DL program.
| Python classes | Protobuf messages | <table>
| --- | --- | <thead>
| Program | ProgramDesc | <tr>
| Block | BlockDesc | <th>Python classes</th>
| Operator | OpDesc | <th>Protobuf messages</th>
| Variable | VarDesc | </tr>
</thead>
<tbody>
<tr>
<td>Program </td>
<td>ProgramDesc </td>
</tr>
<tr>
<td>Block </td>
<td>BlockDesc </td>
</tr>
<tr>
<td>Operator </td>
<td>OpDesc </td>
</tr>
<tr>
<td>Variable </td>
<td>VarDesc </td>
</tr>
</tbody>
</table>
Please be aware that these Python classes need to maintain some construction-time information, which are not part of the protobuf messages. Please be aware that these Python classes need to maintain some construction-time information, which are not part of the protobuf messages.
......
doc/fluid/design/motivation/fluid.md
浏览文件 @ 1f6e0448
...@@ -10,11 +10,37 @@ Fluid is the answer. Fluid is similar to PyTorch and TensorFlow Eager Execution ...@@ -10,11 +10,37 @@ Fluid is the answer. Fluid is similar to PyTorch and TensorFlow Eager Execution
Deep learning infrastructure is one of the fastest evolving technologies. Within four years, there have already been three generations of technologies invented. Deep learning infrastructure is one of the fastest evolving technologies. Within four years, there have already been three generations of technologies invented.
| Existed since | model as sequence of layers | model as graph of operators | No model | <table>
|--|--|--|--| <thead>
| 2013 | Caffe, Theano, Torch, PaddlePaddle | | | <tr>
| 2015 | | TensorFlow, MxNet, Caffe2, ONNX, n-graph | | <th>Existed since</th>
| 2016 | | | PyTorch, TensorFlow Eager Execution, PaddlePaddle Fluid | <th>model as sequence of layers</th>
<th>model as graph of operators</th>
<th>No model</th>
</tr>
</thead>
<tbody>
<tr>
<td>2013 </td>
<td>Caffe, Theano, Torch, PaddlePaddle </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td>2015 </td>
<td> </td>
<td>TensorFlow, MxNet, Caffe2, ONNX, n-graph </td>
<td> </td>
</tr>
<tr>
<td>2016 </td>
<td> </td>
<td> </td>
<td> PyTorch, TensorFlow Eager Execution, PaddlePaddle Fluid</td>
</tr>
</tbody>
</table>
From the above table, we see that the deep learning technology is evolving towards getting rid of the concept of a model. To understand the reasons behind this direction, a comparison of the *programming paradigms* or the ways to program deep learning applications using these systems, would be helpful. The following section goes over these. From the above table, we see that the deep learning technology is evolving towards getting rid of the concept of a model. To understand the reasons behind this direction, a comparison of the *programming paradigms* or the ways to program deep learning applications using these systems, would be helpful. The following section goes over these.
......
doc/fluid/design/motivation/refactorization.md
浏览文件 @ 1f6e0448
...@@ -36,11 +36,37 @@ At compile time, the Python program generates a protobuf message representation ...@@ -36,11 +36,37 @@ At compile time, the Python program generates a protobuf message representation
At runtime, the C++ program realizes the graph and runs it. At runtime, the C++ program realizes the graph and runs it.
| | Representation (protobuf messages) | Realization (C++ class objects) | <table>
|---|---|---| <thead>
|Data|[VarDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L107)|[Variable](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/variable.h#L24)| <tr>
|Operation|[OpDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L35)|[Operator](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/operator.h#L64)| <th></th>
|Block|BlockDesc|Block| <th>Representation (protobuf messages)</th>
<th>Realization (C++ class objects) </th>
</tr>
</thead>
<tbody>
<tr>
<td>Data</td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L107">VarDesc</a></td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/variable.h#L24">Variable</a></td>
</tr>
<tr>
<td>Operation </td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L35">OpDesc</a></td>
<td>
<a href="https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/operator.h#L64">Operator</a></td>
</tr>
<tr>
<td>Block </td>
<td>BlockDesc </td>
<td>Block </td>
</tbody>
</table>
The word *graph* is interchangeable with *block* in this document. A graph consists of computation steps and local variables similar to a C++/Java program block, or a pair of parentheses(`{` and `}`). The word *graph* is interchangeable with *block* in this document. A graph consists of computation steps and local variables similar to a C++/Java program block, or a pair of parentheses(`{` and `}`).
......
doc/fluid/design/network/deep_speech_2.md
浏览文件 @ 1f6e0448
# DeepSpeech2 on PaddlePaddle: Design Doc # DeepSpeech2 on PaddlePaddle: Design Doc
We are planning to build Deep Speech 2 (DS2) \[[1](#references)\], a powerful Automatic Speech Recognition (ASR) engine, on PaddlePaddle. For the first-stage plan, we have the following short-term goals: We are planning to build Deep Speech 2 (DS2) \[[1](#references)\], a powerful Automatic Speech Recognition (ASR) engine, on PaddlePaddle. For the first-stage plan, we have the following short-term goals:
...@@ -68,11 +68,33 @@ We roughly break down the project into 14 tasks: ...@@ -68,11 +68,33 @@ We roughly break down the project into 14 tasks:
Tasks parallelizable within phases: Tasks parallelizable within phases:
Roadmap | Description | Parallelizable Tasks <table>
----------- | :------------------------------------ | :-------------------- <thead>
Phase I | Simplified model & components | *Task 1* ~ *Task 8* <tr>
Phase II | Standard model & benchmarking & profiling | *Task 9* ~ *Task 12* <th>Roadmap</th>
Phase III | Documentations | *Task13* ~ *Task14* <th>Description</th>
<th> Parallelizable Tasks</th>
</tr>
</thead>
<tbody>
<tr>
<td>Phase I </td>
<td>Simplified model & components </td>
<td>Task 1 ~ Task 8</td>
</tr>
<tr>
<td>Phase II </td>
<td> Standard model & benchmarking & profiling</td>
<td>Task 9 ~ Task 12 </td>
</tr>
<tr>
<td>Phase III </td>
<td> Documentations</td>
<td> Task13 ~ Task14 </td>
</tr>
</tbody>
</table>
Issue for each task will be created later. Contributions, discussions and comments are all highly appreciated and welcomed! Issue for each task will be created later. Contributions, discussions and comments are all highly appreciated and welcomed!
...@@ -102,37 +124,82 @@ We don't have to persist on this 2-3-7-1-1-1 depth \[[2](#references)\]. Similar ...@@ -102,37 +124,82 @@ We don't have to persist on this 2-3-7-1-1-1 depth \[[2](#references)\]. Similar
Key ingredients about the layers: Key ingredients about the layers:
- **Data Layers**: - **Data Layers**:
- Frame sequences data of audio **spectrogram** (with FFT). - Frame sequences data of audio **spectrogram** (with FFT).
- Token sequences data of **transcription** text (labels). - Token sequences data of **transcription** text (labels).
- These two type of sequences do not have the same lengthes, thus a CTC-loss layer is required. - These two type of sequences do not have the same lengthes, thus a CTC-loss layer is required.
- **2D Convolution Layers**: - **2D Convolution Layers**:
- Not only temporal convolution, but also **frequency convolution**. Like a 2D image convolution, but with a variable dimension (i.e. temporal dimension). - Not only temporal convolution, but also **frequency convolution**. Like a 2D image convolution, but with a variable dimension (i.e. temporal dimension).
- With striding for only the first convlution layer. - With striding for only the first convlution layer.
- No pooling for all convolution layers. - No pooling for all convolution layers.
- **Uni-directional RNNs** - **Uni-directional RNNs**
- Uni-directional + row convolution: for low-latency inference. - Uni-directional + row convolution: for low-latency inference.
- Bi-direcitional + without row convolution: if we don't care about the inference latency. - Bi-direcitional + without row convolution: if we don't care about the inference latency.
- **Row convolution**: - **Row convolution**:
- For looking only a few steps ahead into the feature, instead of looking into a whole sequence in bi-directional RNNs. - For looking only a few steps ahead into the feature, instead of looking into a whole sequence in bi-directional RNNs.
- Not nessesary if with bi-direcitional RNNs. - Not nessesary if with bi-direcitional RNNs.
- "**Row**" means convolutions are done within each frequency dimension (row), and no convolution kernels shared across. - "**Row**" means convolutions are done within each frequency dimension (row), and no convolution kernels shared across.
- **Batch Normalization Layers**: - **Batch Normalization Layers**:
- Added to all above layers (except for data and loss layer). - Added to all above layers (except for data and loss layer).
- Sequence-wise normalization for RNNs: BatchNorm only performed on input-state projection and not state-state projection, for efficiency consideration. - Sequence-wise normalization for RNNs: BatchNorm only performed on input-state projection and not state-state projection, for efficiency consideration.
<table>
Required Components | PaddlePaddle Support | Need to Develop <thead>
:------------------------------------- | :-------------------------------------- | :----------------------- <tr>
Data Layer I (Spectrogram) | Not supported yet. | TBD (Task 3) <th>Required Components</th>
Data Layer II (Transcription) | `paddle.data_type.integer_value_sequence` | - <th> PaddlePaddle Support</th>
2D Convolution Layer | `paddle.layer.image_conv_layer` | - <th> Need to Develop</th>
DataType Converter (vec2seq) | `paddle.layer.block_expand` | - </tr>
Bi-/Uni-directional RNNs | `paddle.layer.recurrent_group` | - </thead>
Row Convolution Layer | Not supported yet. | TBD (Task 4) <tbody>
CTC-loss Layer | `paddle.layer.warp_ctc` | - <tr>
Batch Normalization Layer | `paddle.layer.batch_norm` | - <td>Data Layer I (Spectrogram) </td>
CTC-Beam search | Not supported yet. | TBD (Task 6) <td>Not supported yet.</td>
<td>TBD (Task 3)</td>
</tr>
<tr>
<td>Data Layer II (Transcription) </td>
<td> paddle.data_type.integer_value_sequence</td>
<td> - </td>
</tr>
<tr>
<td>2D Convolution Layer </td>
<td> paddle.layer.image_conv_layer</td>
<td> - </td>
</tr>
<tr>
<td>DataType Converter (vec2seq)</td>
<td> paddle.layer.block_expand</td>
<td> - </td>
</tr>
<tr>
<td>Bi-/Uni-directional RNNs </td>
<td>paddle.layer.recurrent_group</td>
<td> - </td>
</tr>
<tr>
<td>Row Convolution Layer </td>
<td>Not supported yet.</td>
<td>TBD (Task 4)</td>
</tr>
<tr>
<td>CTC-loss Layer </td>
<td>paddle.layer.warp_ctc</td>
<td> - </td>
</tr>
<tr>
<td>Batch Normalization Layer </td>
<td>paddle.layer.batch_norm</td>
<td> - </td>
</tr>
<tr>
<td>CTC-Beam search </td>
<td>Not supported yet.</td>
<td> TBD (Task 6) </td>
</tr>
</tbody>
</table>
### Row Convolution ### Row Convolution
...@@ -145,14 +212,14 @@ TODO by Assignees ...@@ -145,14 +212,14 @@ TODO by Assignees
Figure 2. Algorithm for CTC Beam Search Decoder. Figure 2. Algorithm for CTC Beam Search Decoder.
</div> </div>
- The **Beam Search Decoder** for DS2 CTC-trained network follows the similar approach in \[[3](#references)\] as shown in Figure 2, with two important modifications for the ambiguous parts: - The **Beam Search Decoder** for DS2 CTC-trained network follows the similar approach in \[[3](#references)\] as shown in Figure 2, with two important modifications for the ambiguous parts:
- 1) in the iterative computation of probabilities, the assignment operation is changed to accumulation for one prefix may comes from different paths; - 1) in the iterative computation of probabilities, the assignment operation is changed to accumulation for one prefix may comes from different paths;
- 2) the if condition ```if l^+ not in A_prev then``` after probabilities' computation is deprecated for it is hard to understand and seems unnecessary. - 2) the if condition ```if l^+ not in A_prev then``` after probabilities' computation is deprecated for it is hard to understand and seems unnecessary.
- An **external scorer** would be passed into the decoder to evaluate a candidate prefix during decoding whenever a white space appended in English decoding and any character appended in Mandarin decoding. - An **external scorer** would be passed into the decoder to evaluate a candidate prefix during decoding whenever a white space appended in English decoding and any character appended in Mandarin decoding.
- Such external scorer consists of language model, word count or any other custom scorers. - Such external scorer consists of language model, word count or any other custom scorers.
- The **language model** is built from Task 5, with parameters should be carefully tuned to achieve minimum WER/CER (c.f. Task 7) - The **language model** is built from Task 5, with parameters should be carefully tuned to achieve minimum WER/CER (c.f. Task 7)
- This decoder needs to perform with **high efficiency** for the convenience of parameters tuning and speech recognition in reality. - This decoder needs to perform with **high efficiency** for the convenience of parameters tuning and speech recognition in reality.
## Future Work ## Future Work
......
doc/fluid/dev/index_cn.rst
浏览文件 @ 1f6e0448
...@@ -4,9 +4,9 @@ ...@@ -4,9 +4,9 @@
.. toctree:: .. toctree::
:maxdepth: 1 :maxdepth: 1
new_op_en.md new_op_cn.md
new_op_kernel_en.md new_op_kernel.md
use_eigen_en.md use_eigen_cn.md
name_convention.md name_convention.md
support_new_device.md support_new_device.md
releasing_process.md releasing_process.md
......
doc/fluid/dev/index_en.rst
浏览文件 @ 1f6e0448
...@@ -5,7 +5,7 @@ Development ...@@ -5,7 +5,7 @@ Development
:maxdepth: 1 :maxdepth: 1
new_op_en.md new_op_en.md
new_op_kernel_en.md new_op_kernel.md
use_eigen_en.md use_eigen_en.md
name_convention.md name_convention.md
support_new_device.md support_new_device.md
......
doc/fluid/dev/new_op_cn.md
浏览文件 @ 1f6e0448
...@@ -26,13 +26,32 @@ ...@@ -26,13 +26,32 @@
依据是否包含kernel，可以将Op分为两种：包含Kernel的Op和不包含kernel的Op，前者Op的定义继承自`OperatorWithKernel`，后者继承自`OperatorBase`。本教程主要介绍带Kernel的Op如何写，简单总结Op需要包含的内容如下： 依据是否包含kernel，可以将Op分为两种：包含Kernel的Op和不包含kernel的Op，前者Op的定义继承自`OperatorWithKernel`，后者继承自`OperatorBase`。本教程主要介绍带Kernel的Op如何写，简单总结Op需要包含的内容如下：
<table>
内容 | 定义位置 <thead>
-------------- | :---------------------- <tr>
OpProtoMake定义 | `.cc`文件，Backward Op不需要定义OpProtoMake <th>内容</th>
Op定义 | `.cc`文件 <th>定义位置</th>
Kernel实现 | CPU、CUDA共享Kernel实现在`.h`文件中，否则，CPU 实现在`.cc`文件中，CUDA 实现在`.cu`文件中。 </tr>
注册Op | Op注册实现在`.cc`文件；Kernel注册CPU实现在`.cc`文件中，CUDA实现在`.cu`文件中 </thead>
<tbody>
<tr>
<td>OpProtoMake定义 </td>
<td>`.cc`文件，Backward Op不需要定义OpProtoMake </td>
</tr>
<tr>
<td>Op定义 </td>
<td> `.cc`文件</td>
</tr>
<tr>
<td>Kernel实现 </td>
<td> CPU、CUDA共享Kernel实现在`.h`文件中，否则，CPU 实现在`.cc`文件中，CUDA 实现在`.cu`文件中。</td>
</tr>
<tr>
<td>注册Op </td>
<td> Op注册实现在`.cc`文件；Kernel注册CPU实现在`.cc`文件中，CUDA实现在`.cu`文件中</td>
</tr>
</tbody>
</table>
实现新的op都添加至目录[paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)下，文件命名以`*_op.h`（如有） 、 `*_op.cc` 、`*_op.cu`（如有）结尾。**系统会根据文件名自动构建op和其对应的Python扩展。** 实现新的op都添加至目录[paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators)下，文件命名以`*_op.h`（如有） 、 `*_op.cc` 、`*_op.cu`（如有）结尾。**系统会根据文件名自动构建op和其对应的Python扩展。**
......
doc/fluid/dev/new_op_en.md
浏览文件 @ 1f6e0448
...@@ -33,6 +33,33 @@ Op definition | `.cc` files ...@@ -33,6 +33,33 @@ Op definition | `.cc` files
Kernel implementation | The kernel methods shared between CPU and CUDA are defined in `.h` files. CPU-specific kernels live in `.cc` files, while CUDA-specific kernels are implemented in `.cu`files. Kernel implementation | The kernel methods shared between CPU and CUDA are defined in `.h` files. CPU-specific kernels live in `.cc` files, while CUDA-specific kernels are implemented in `.cu`files.
Registering the Op | Ops are registered in `.cc` files; For Kernel registration, `.cc` files contain the CPU implementation, while `.cu` files contain the CUDA implementation. Registering the Op | Ops are registered in `.cc` files; For Kernel registration, `.cc` files contain the CPU implementation, while `.cu` files contain the CUDA implementation.
<table>
<thead>
<tr>
<th>Information</th>
<th> Where is it defined</th>
</tr>
</thead>
<tbody>
<tr>
<td>OpProtoMake definition </td>
<td> `.cc`files, Backward Op does not need an OpProtoMake interface. </td>
</tr>
<tr>
<td>Op definition </td>
<td> `.cc` files</td>
</tr>
<tr>
<td>Kernel implementation </td>
<td> The kernel methods shared between CPU and CUDA are defined in `.h` files. CPU-specific kernels live in `.cc` files, while CUDA-specific kernels are implemented in `.cu`files.</td>
</tr>
<tr>
<td>Registering the Op </td>
<td> Ops are registered in `.cc` files; For Kernel registration, `.cc` files contain the CPU implementation, while `.cu` files contain the CUDA implementation.</td>
</tr>
</tbody>
</table>
New Operator implementations are added to the list [paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators), with file names in the format `*_op.h` (if applicable), `*_op.cc`, `*_op.cu` (if applicable).** The system will use the naming scheme to automatically build operators and their corresponding Python extensions.** New Operator implementations are added to the list [paddle/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators), with file names in the format `*_op.h` (if applicable), `*_op.cc`, `*_op.cu` (if applicable).** The system will use the naming scheme to automatically build operators and their corresponding Python extensions.**
...@@ -279,7 +306,7 @@ A forward operator unit test inherits `unittest.TestCase` and defines metaclass ...@@ -279,7 +306,7 @@ A forward operator unit test inherits `unittest.TestCase` and defines metaclass
def test_check_output(self): def test_check_output(self):
self.check_output() self.check_output()
def test_check_grad_normal(self): def test_check_grad_normal(self):
self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5) self.check_grad(['X', 'Y'], 'Out', max_relative_error=0.5)
......
doc/fluid/dev/releasing_process.md
浏览文件 @ 1f6e0448
...@@ -66,7 +66,7 @@ PaddlePaddle开发过程使用[git-flow](http://nvie.com/posts/a-successful-git- ...@@ -66,7 +66,7 @@ PaddlePaddle开发过程使用[git-flow](http://nvie.com/posts/a-successful-git-
* 建议，开发者fork的版本库使用`develop`分支同步主版本库的`develop`分支 * 建议，开发者fork的版本库使用`develop`分支同步主版本库的`develop`分支
* 建议，开发者fork的版本库中，再基于`develop`版本fork出自己的功能分支。 * 建议，开发者fork的版本库中，再基于`develop`版本fork出自己的功能分支。
* 当功能分支开发完毕后，向PaddlePaddle的主版本库提交`Pull Reuqest`，进而进行代码评审。 * 当功能分支开发完毕后，向PaddlePaddle的主版本库提交`Pull Reuqest`，进而进行代码评审。
* 在评审过程中，开发者修改自己的代码，可以继续在自己的功能分支提交代码。 * 在评审过程中，开发者修改自己的代码，可以继续在自己的功能分支提交代码。
* BugFix分支也是在开发者自己的fork版本库维护，与功能分支不同的是，BugFix分支需要分别给主版本库的`master`、`develop`与可能有的`release/版本号`分支，同时提起`Pull Request`。 * BugFix分支也是在开发者自己的fork版本库维护，与功能分支不同的是，BugFix分支需要分别给主版本库的`master`、`develop`与可能有的`release/版本号`分支，同时提起`Pull Request`。
...@@ -78,13 +78,116 @@ PaddlePaddle开发过程使用[git-flow](http://nvie.com/posts/a-successful-git- ...@@ -78,13 +78,116 @@ PaddlePaddle开发过程使用[git-flow](http://nvie.com/posts/a-successful-git-
PaddlePaddle每次发版本首先要保证PaddlePaddle Book中所有章节功能的正确性。功能的正确性包括验证PaddlePaddle目前的`paddle_trainer`训练和纯使用`Python`训练模型正确性。 PaddlePaddle每次发版本首先要保证PaddlePaddle Book中所有章节功能的正确性。功能的正确性包括验证PaddlePaddle目前的`paddle_trainer`训练和纯使用`Python`训练模型正确性。
| | 新手入门章节 | 识别数字 | 图像分类 | 词向量 | 情感分析 | 语意角色标注 | 机器翻译 | 个性化推荐 | <table>
| --- | --- | --- | --- | --- | --- | --- | --- | --- | <thead>
| API.V2 + Docker + GPU | | | | | | | | | <tr>
| API.V2 + Docker + CPU | | | | | | | | | <th></th>
| `paddle_trainer` + Docker + GPU | | | | | | | | | <th>新手入门章节 </th>
| `paddle_trainer` + Docker + CPU | | | | | | | | | <th> 识别数字</th>
| API.V2 + Ubuntu + GPU | | | | | | | | | <th> 图像分类</th>
| API.V2 + Ubuntu + CPU | | | | | | | | | <th>词向量</th>
| `paddle_trainer` + Ubuntu + GPU | | | | | | | | | <th> 情感分析</th>
| `paddle_trainer` + Ubuntu + CPU | | | | | | | | | <th>语意角色标注</th>
<th> 机器翻译</th>
<th>个性化推荐</th>
</tr>
</thead>
<tbody>
<tr>
<td>API.V2 + Docker + GPU </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td> API.V2 + Docker + CPU </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td>`paddle_trainer` + Docker + GPU </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td>`paddle_trainer` + Docker + CPU </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td> API.V2 + Ubuntu + GPU</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td>API.V2 + Ubuntu + CPU </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td> `paddle_trainer` + Ubuntu + GPU</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr>
<td> `paddle_trainer` + Ubuntu + CPU</td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
</tbody>
</table>
doc/fluid/getstarted/concepts/save_model/model_format.md
浏览文件 @ 1f6e0448
...@@ -4,30 +4,70 @@ ...@@ -4,30 +4,70 @@
A model is an output of the training process. One complete model consists of two parts, the **topology** and the **parameters**. In order to support industrial deployment, the model format must be self-complete and must not expose any training source code. A model is an output of the training process. One complete model consists of two parts, the **topology** and the **parameters**. In order to support industrial deployment, the model format must be self-complete and must not expose any training source code.
As a result, In PaddlePaddle, the **topology** is represented as a [ProgramDesc](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/doc/design/program.md), which describes the model structure. The **parameters** contain all the trainable weights in the model. We must support large size parameters and efficient serialization/deserialization of parameters. As a result, In PaddlePaddle, the **topology** is represented as a [ProgramDesc](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/doc/design/program.md), which describes the model structure. The **parameters** contain all the trainable weights in the model. We must support large size parameters and efficient serialization/deserialization of parameters.
## Implementation ## Implementation
The topology is saved as a plain text in a detailed self-contain protobuf file. The topology is saved as a plain text in a detailed self-contain protobuf file.
The parameters are saved as a binary file. As we all know, the protobuf message has a limit of [64M size](https://developers.google.com/protocol-buffers/docs/reference/cpp/google.protobuf.io.coded_stream#CodedInputStream.SetTotalBytesLimit.details). We have done a [benchmark experiment](https://github.com/PaddlePaddle/Paddle/pull/4610), which shows that protobuf is not fit for the task. The parameters are saved as a binary file. As we all know, the protobuf message has a limit of [64M size](https://developers.google.com/protocol-buffers/docs/reference/cpp/google.protobuf.io.coded_stream#CodedInputStream.SetTotalBytesLimit.details). We have done a [benchmark experiment](https://github.com/PaddlePaddle/Paddle/pull/4610), which shows that protobuf is not fit for the task.
As a result, we design a particular format for tensor serialization. By default, an arbitrary tensor in Paddle is a [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor.md), and has a description information proto of [LoDTensorDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L99). We save the DescProto as the byte string header. It contains all the necessary information, such as the `dims`, and the `LoD` information in [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/paddle/framework/lod_tensor.md). A tensor stores values in a continuous memory buffer. For speed we dump the raw memory to disk and save it as the byte string content. So, the binary format of one tensor is, As a result, we design a particular format for tensor serialization. By default, an arbitrary tensor in Paddle is a [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/lod_tensor.md), and has a description information proto of [LoDTensorDesc](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/framework.proto#L99). We save the DescProto as the byte string header. It contains all the necessary information, such as the `dims`, and the `LoD` information in [LoDTensor](https://github.com/PaddlePaddle/Paddle/blob/1c0a4c901c9fc881d120249c703b15d1c50dae7d/paddle/framework/lod_tensor.md). A tensor stores values in a continuous memory buffer. For speed we dump the raw memory to disk and save it as the byte string content. So, the binary format of one tensor is,
The table below shows a tensor's byte view in detail. Note that all the signed values are written in the little-endian format. The table below shows a tensor's byte view in detail. Note that all the signed values are written in the little-endian format.
|field name | type | description | <table>
| --- | --- | --- | <thead>
| version | uint32_t | Version of saved file. Always 0 now. | <tr>
| tensor desc length | uint32_t | TensorDesc(Protobuf message) length in bytes. | <th>field name</th>
| tensor desc | void* | TensorDesc protobuf binary message | <th>type </th>
| tensor data | void* | Tensor's data in binary format. The length of `tensor_data` is decided by `TensorDesc.dims()` and `TensorDesc.data_type()` | <th>description </th>
| lod_level | uint64_t | Level of LoD | </tr>
| length of lod[0] | uint64_t | [Optional] length of lod[0] in bytes. | </thead>
| data of lod[0] | uint64_t* | [Optional] lod[0].data() | <tbody>
| ... | ... | ... | <tr>
<td> version</td>
<td> uint32_t </td>
<td> Version of saved file. Always 0 now.</td>
</tr>
<tr>
<td> tensor desc length </td>
<td> uint32_t </td>
<td> TensorDesc(Protobuf message) length in bytes. </td>
</tr>
<tr>
<td>tensor desc </td>
<td> void*</td>
<td> TensorDesc protobuf binary message </td>
</tr>
<tr>
<td> tensor data </td>
<td> void* </td>
<td> Tensor's data in binary format. The length of `tensor_data` is decided by `TensorDesc.dims()` and `TensorDesc.data_type()` </td>
</tr>
<tr>
<td> lod_level</td>
<td> uint64_t </td>
<td> Level of LoD </td>
</tr>
<tr>
<td> length of lod[0] </td>
<td> uint64_t </td>
<td> [Optional] length of lod[0] in bytes. </td>
</tr>
<tr>
<td> data of lod[0] </td>
<td> uint64_t* </td>
<td> [Optional] lod[0].data() </td>
</tr>
<tr>
<td>... </td>
<td> ... </td>
<td> ... </td>
</tr>
</tbody>
</table>
## Summary ## Summary
......
doc/fluid/howto/cluster/fluid_cluster_train_cn.md
浏览文件 @ 1f6e0448
...@@ -65,10 +65,10 @@ exit(1) ...@@ -65,10 +65,10 @@ exit(1)
**因此，在分布式的Fluid环境中，我们有两个角色需要创建，分别是Parameter Server和Trainer。** **因此，在分布式的Fluid环境中，我们有两个角色需要创建，分别是Parameter Server和Trainer。**
### 分布式训练 ### 分布式训练
Fliud专门提供了工具[Distributed Transpiler](https://github.com/PaddlePaddle/Paddle/blob/ba65d54d9d3b41cd3c5171b00f476d4e60133ddb/doc/fluid/design/dist_train/distributed_architecture.md#distributed-transpiler)用于将单机版的训练程序转换为分布式版本的训练程序。工具背后的理念是找出程序的优化算子和梯度参数，将他们分隔为两部分，通过send/recv 操作算子进行连接,优化算子和梯度参数可以在优化器的minimize函数的返回值中获取到。 Fliud专门提供了工具[Distributed Transpiler](https://github.com/PaddlePaddle/Paddle/blob/ba65d54d9d3b41cd3c5171b00f476d4e60133ddb/doc/fluid/design/dist_train/distributed_architecture.md#distributed-transpiler)用于将单机版的训练程序转换为分布式版本的训练程序。工具背后的理念是找出程序的优化算子和梯度参数，将他们分隔为两部分，通过send/recv 操作算子进行连接,优化算子和梯度参数可以在优化器的minimize函数的返回值中获取到。
```python ```python
optimize_ops, params_grads = sgd_optimizer.minimize(avg_cost) optimize_ops, params_grads = sgd_optimizer.minimize(avg_cost)
``` ```
将Distributed Transpiler、优化算子和梯度函数放在一个代码中如下： 将Distributed Transpiler、优化算子和梯度函数放在一个代码中如下：
```python ```python
...@@ -99,15 +99,51 @@ for pass_id in range(100): ...@@ -99,15 +99,51 @@ for pass_id in range(100):
### 分布式训练脚本运行说明 ### 分布式训练脚本运行说明
分布式任务的运行需要将表格中说明的多个参数进行赋值: 分布式任务的运行需要将表格中说明的多个参数进行赋值:
| 参数名 | 值类型 | 说明 | 示例 | <table>
|:-------------|:------|:---------------------------------------|:-------------| <thead>
| trainer_id | int | 当前训练节点的ID，训练节点ID编号为0 - n-1， n为trainers的值 | 0/1/2/3 | <tr>
| pservers | str | parameter server 列表 | 127.0.0.1:6710,127.0.0.1:6711 | <th>参数名</th>
| trainers | int | 训练节点的总个数，>0的数字 | 4 | <th> 值类型</th>
| server_endpoint | str | 当前所起的服务节点的IP:PORT | 127.0.0.1:8789 | <th>说明</th>
| training_role | str | 节点角色， TRAINER/PSERVER | PSERVER | <th> 示例</th>
</tr>
**注意：** ```training_role```是用来区分当前所起服务的角色的，用于训练程序中，用户可根据需要自行定义，其他参数为fluid.DistributeTranspiler的transpile函数所需要，需要在调用函数前进行定义，样例如下： </thead>
<tbody>
<tr>
<td>trainer_id </td>
<td> int</td>
<td> 当前训练节点的ID，训练节点ID编号为0 - n-1， n为trainers的值 </td>
<td> 0/1/2/3 </td>
</tr>
<tr>
<td>pservers </td>
<td> str</td>
<td> parameter server 列表 </td>
<td> 127.0.0.1:6710,127.0.0.1:6711 </td>
</tr>
<tr>
<td>trainers </td>
<td>int </td>
<td> 训练节点的总个数，>0的数字 </td>
<td> 4 </td>
</tr>
<tr>
<td> server_endpoint</td>
<td> str </td>
<td> 当前所起的服务节点的IP:PORT </td>
<td> 127.0.0.1:8789 </td>
</tr>
<tr>
<td> training_role</td>
<td>str </td>
<td> 节点角色， TRAINER/PSERVER </td>
<td> PSERVER </td>
</tr>
</tbody>
</table>
**注意：** ```training_role```是用来区分当前所起服务的角色的，用于训练程序中，用户可根据需要自行定义，其他参数为fluid.DistributeTranspiler的transpile函数所需要，需要在调用函数前进行定义，样例如下：
```python ```python
t = fluid.DistributeTranspiler() t = fluid.DistributeTranspiler()
......
doc/fluid/howto/optimization/cpu_profiling_cn.md
浏览文件 @ 1f6e0448
...@@ -42,14 +42,40 @@ cprofilev -a 0.0.0.0 -p 3214 -f profile.out main.py ...@@ -42,14 +42,40 @@ cprofilev -a 0.0.0.0 -p 3214 -f profile.out main.py
每一列的含义是: 每一列的含义是:
| 列名 | 含义 | <table>
| --- | --- | <thead>
| ncalls | 函数的调用次数 | <tr>
| tottime | 函数实际使用的总时间。该时间去除掉本函数调用其他函数的时间 | <th>列名</th>
| percall | tottime的每次调用平均时间 | <th>含义 </th>
| cumtime | 函数总时间。包含这个函数调用其他函数的时间 | </tr>
| percall | cumtime的每次调用平均时间 | </thead>
| filename:lineno(function) | 文件名, 行号，函数名 | <tbody>
<tr>
<td> ncalls</td>
<td> 函数的调用次数</td>
</tr>
<tr>
<td>tottime</td>
<td> 函数实际使用的总时间。该时间去除掉本函数调用其他函数的时间</td>
</tr>
<tr>
<td> percall </td>
<td> tottime的每次调用平均时间</td>
</tr>
<tr>
<td> cumtime</td>
<td> 函数总时间。包含这个函数调用其他函数的时间</td>
</tr>
<tr>
<td> percall</td>
<td> cumtime的每次调用平均时间</td>
</tr>
<tr>
<td> filename:lineno(function) </td>
<td> 文件名, 行号，函数名 </td>
</tr>
</tbody>
</table>
### 寻找性能瓶颈 ### 寻找性能瓶颈
......
doc/fluid/howto/optimization/cpu_profiling_en.md
浏览文件 @ 1f6e0448
...@@ -57,14 +57,40 @@ port, we will see the output like the following: ...@@ -57,14 +57,40 @@ port, we will see the output like the following:
where each line corresponds to Python function, and the meaning of where each line corresponds to Python function, and the meaning of
each column is as follows: each column is as follows:
| column | meaning | <table>
| --- | --- | <thead>
| ncalls | the number of calls into a function | <tr>
| tottime | the total execution time of the function, not including the execution time of other functions called by the function | <th>column</th>
| percall | tottime divided by ncalls | <th>meaning </th>
| cumtime | the total execution time of the function, including the execution time of other functions being called | </tr>
| percall | cumtime divided by ncalls | </thead>
| filename:lineno(function) | where the function is defined | <tbody>
<tr>
<td> ncalls</td>
<td> the number of calls into a function</td>
</tr>
<tr>
<td>tottime</td>
<td> the total execution time of the function, not including the execution time of other functions called by the function</td>
</tr>
<tr>
<td> percall </td>
<td> tottime divided by ncalls</td>
</tr>
<tr>
<td> cumtime</td>
<td> the total execution time of the function, including the execution time of other functions being called</td>
</tr>
<tr>
<td> percall</td>
<td> cumtime divided by ncalls</td>
</tr>
<tr>
<td> filename:lineno(function) </td>
<td> where the function is define </td>
</tr>
</tbody>
</table>
### Identify Performance Bottlenecks ### Identify Performance Bottlenecks
......
doc/v2/faq/build_and_install/index_en.rst
浏览文件 @ 1f6e0448
############################ .. _install_faq:
Install, Build and Unit test
############################
TBD ###############################
Compile, Install, and Unit Test
###############################
.. contents::
1. Insufficient CUDA driver version
----------------------------------------------------------------
Many users usually face issues like `Cuda Error: CUDA driver version is insufficient for CUDA runtime version` when running the PaddlePaddle GPU Docker image. The cause is that you may not map the local CUDA driver to a container directory.
You can solve the issue by running the following commands:
.. code-block:: bash
$ export CUDA_SO="$(\ls usr/lib64/libcuda* | xargs -I{} echo '-v {}:{}') $(\ls /usr/lib64/libnvidia* | xargs -I{} echo '-v {}:{}')"
$ export DEVICES=$(\ls /dev/nvidia* | xargs -I{} echo '--device {}:{}')
$ docker run ${CUDA_SO} ${DEVICES} -it paddlepaddle/paddle:latest-gpu
For more infomation about Docker's installation and usage, please refer to `PaddlePaddle Docker documentation <http://www.paddlepaddle.org/docs/0.11.0/documentation/zh/getstarted/build_and_install/docker_install_en.html>`_ .
2. Version mismatch between PythonLibs and PythonInterpreter
----------------------------------------------------------------
It is a common bug when CMake looks up Python. If you install multiple versions of Python, Cmake may find the version mismatch between PythonLibs and PythonInterpreter . You are forced to specify a Python version, as follows.
.. code-block:: bash
cmake .. -DPYTHON_EXECUTABLE=<exc_path> -DPYTHON_LIBRARY=<lib_path> -DPYTHON_INCLUDE_DIR=<inc_path>
You should specify ``<exc_path>``, ``<lib_path>``, ``<inc_path>`` to your local paths.
3. PaddlePaddle version is 0.0.0
------------------------------------------------
This issue would happen when you run the code `paddle version` or `cmake ..`
.. code-block:: bash
CMake Warning at cmake/version.cmake:20 (message):
Cannot add paddle version from git tag
You should pull all remote branches to your local machine with the command :code:`git fetch upstream` and then run :code:`cmake`
4. paddlepaddle\*.whl is not a supported wheel on this platform.
------------------------------------------------------------------------
The primary cause for this issue is that it can not find the correct PaddlePaddle installation package that matches your current system.The latest PaddlePaddle Python installation package supports Linux x86_64 and MacOS 10.12 os including Python2.7 and Pip 9.0.1.
You can upgrade Pip with the following command\:
.. code-block:: bash
pip install --upgrade pip
If it does not work for you, you can run the command :code:`python -c "import pip; print(pip.pep425tags.get_supported())"` to get the suffix of Python package which your system may support and then compare it with the suffix of your installation.
If the system supports :code:`linux_x86_64` and the installation package is :code:`manylinux1_x86_64`, you should upgrade pip to the latest
if the system supports :code:`manylinux_x86_64` and the local installation package is :code:`linux1_x86_64`, you can rename the whl package to :code:`manylinux1_x86_64` and then try again.
5. ImportError: No module named v2
----------------------------------
Please uninstall Paddle V1 if you have installed it before.
.. code-block:: bash
pip uninstall py_paddle paddle
Then install Python for PaddlePaddle , enter the build directory and run the following commands
pip install python/dist/paddle*.whl && pip install ../paddle/dist/py_paddle*.whl
6. Illegal instruction
-----------------------
This issue may be caused by the wrong usage of PaddlePaddle binary version which uses avx SIMD instructions to increase the performance of cpu. Please choose the correct version.
7. Python unittest fails
--------------------------------
If the following python unittest testcases fail:
.. code-block:: bash
24 - test_PyDataProvider (Failed)
26 - test_RecurrentGradientMachine (Failed)
27 - test_NetworkCompare (Failed)
28 - test_PyDataProvider2 (Failed)
32 - test_Prediction (Failed)
33 - test_Compare (Failed)
34 - test_Trainer (Failed)
35 - test_TrainerOnePass (Failed)
36 - test_CompareTwoNets (Failed)
37 - test_CompareTwoOpts (Failed)
38 - test_CompareSparse (Failed)
39 - test_recurrent_machine_generation (Failed)
40 - test_PyDataProviderWrapper (Failed)
41 - test_config_parser (Failed)
42 - test_swig_api (Failed)
43 - layers_test (Failed)
Please check the PaddlePaddle unittest logs which may suggest the following:
.. code-block:: bash
paddle package is already in your PYTHONPATH. But unittest need a clean environment.
Please uninstall paddle package before start unittest. Try to 'pip uninstall paddle'.
The solution is:
* Remove old PaddlePaddle to make a clean environment for the unit tests. If PaddlePaddle package is already in Python's site-packages, unit tests would refer Python package in site-packages instead of Python package in the :code:`/python` directory of the source directory. Setting :code:`PYTHONPATH` to :code:`/python` is also useless because Python's search path would give the priority to the installed Python package.
8. Failed to download the MKLML library
----------------------------------------------
.. code-block:: bash
make[2]: *** [third_party/mklml/src/extern_mklml-stamp/extern_mklml-download] error 4
make[1]: *** [CMakeFiles/extern_mklml.dir/all] error 2
make[1]: *** waiting for the unfinished jobs....
Cause: The network speed or SSL link causes the MKLML library to download unsuccessfully.
The solution is: manually download and install, the specific steps are as follows.
.. code-block:: bash
// 1. enter the directory
cd build/third_party/mklml/src/extern_mklml
// 2. check the size of the package, normally 75M, if less than 75M, the download fails
du -sh mklml_lnx_2018.0.1.20171007.tgz
// 3. manually download and unzip and make the download success tag:
wget --no-check-certificate https://github.com/01org/mkl-dnn/releases/download/v0.11/mklml_lnx_2018.0.1.20171007.tgz -c -O mklml_lnx_2018.0.1.20171007.tgz
tar zxf mklml_lnx_2018.0.1.20171007.tgz
touch ../extern_mklml-stamp/extern_mklml-download
// 4. then compile
paddle/fluid/framework/details/multi_devices_graph_builder.cc
浏览文件 @ 1f6e0448
...@@ -55,6 +55,7 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build( ...@@ -55,6 +55,7 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build(
const ProgramDesc &program) const { const ProgramDesc &program) const {
auto graph = new SSAGraph(); auto graph = new SSAGraph();
SSAGraph &result = *graph; SSAGraph &result = *graph;
std::unordered_set<std::string> og_has_been_broadcast;
result.vars_.resize(places_.size()); result.vars_.resize(places_.size());
bool is_forwarding = true; bool is_forwarding = true;
...@@ -122,9 +123,15 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build( ...@@ -122,9 +123,15 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build(
if (!is_forwarding) { if (!is_forwarding) {
auto var_names = op->OutputArgumentNames(); auto var_names = op->OutputArgumentNames();
// Currently, we assume that once gradient is generated, it can be
// broadcast, and each gradient is only broadcast once. But there are no
// other cases, for example, we need to adjust the gradient according to
// the input when we get the gradient, which is not considered at present.
for (auto &og : var_names) { for (auto &og : var_names) {
if (grad_names_.count(og) != 0) { // is param grad if (grad_names_.count(og) != 0 &&
// Insert NCCL AllReduce Op og_has_been_broadcast.count(og) == 0) { // is param grad
// Insert NCCL AllReduce Op
og_has_been_broadcast.insert(og);
#ifdef PADDLE_WITH_CUDA #ifdef PADDLE_WITH_CUDA
result.ops_.emplace_back( result.ops_.emplace_back(
new NCCLAllReduceOpHandle(local_scopes_, places_, *nccl_ctxs_)); new NCCLAllReduceOpHandle(local_scopes_, places_, *nccl_ctxs_));
......
paddle/fluid/framework/details/nccl_all_reduce_op_handle.cc
浏览文件 @ 1f6e0448
...@@ -76,7 +76,7 @@ void NCCLAllReduceOpHandle::RunImpl() { ...@@ -76,7 +76,7 @@ void NCCLAllReduceOpHandle::RunImpl() {
} }
} }
std::string NCCLAllReduceOpHandle::Name() const { return "NCCL AllReduce"; } std::string NCCLAllReduceOpHandle::Name() const { return "nccl_all_reduce"; }
} // namespace details } // namespace details
} // namespace framework } // namespace framework
} // namespace paddle } // namespace paddle
paddle/fluid/framework/details/nccl_all_reduce_op_handle.h
浏览文件 @ 1f6e0448
...@@ -14,6 +14,9 @@ ...@@ -14,6 +14,9 @@
#pragma once #pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/details/op_handle_base.h" #include "paddle/fluid/framework/details/op_handle_base.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/scope.h" #include "paddle/fluid/framework/scope.h"
...@@ -34,6 +37,10 @@ struct NCCLAllReduceOpHandle : public OpHandleBase { ...@@ -34,6 +37,10 @@ struct NCCLAllReduceOpHandle : public OpHandleBase {
std::string Name() const override; std::string Name() const override;
// Delay and buffer nccl_all_reduce together can significantly increase
// performance. Disable this feature by returning false.
bool IsMultiDeviceTransfer() override { return true; };
protected: protected:
void RunImpl() override; void RunImpl() override;
}; };
......
paddle/fluid/framework/details/op_handle_base.h
浏览文件 @ 1f6e0448
...@@ -13,6 +13,8 @@ ...@@ -13,6 +13,8 @@
// limitations under the License. // limitations under the License.
#pragma once #pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/details/var_handle.h" #include "paddle/fluid/framework/details/var_handle.h"
#include "paddle/fluid/platform/device_context.h" #include "paddle/fluid/platform/device_context.h"
...@@ -53,6 +55,10 @@ class OpHandleBase { ...@@ -53,6 +55,10 @@ class OpHandleBase {
void AddOutput(VarHandleBase *out); void AddOutput(VarHandleBase *out);
// If the Op involves data transfer of multiple devices that
// will likely block other computations.
virtual bool IsMultiDeviceTransfer() { return false; }
protected: protected:
virtual void RunImpl() = 0; virtual void RunImpl() = 0;
}; };
......
paddle/fluid/framework/details/threaded_ssa_graph_executor.cc
浏览文件 @ 1f6e0448
...@@ -23,22 +23,36 @@ ThreadedSSAGraphExecutor::ThreadedSSAGraphExecutor( ...@@ -23,22 +23,36 @@ ThreadedSSAGraphExecutor::ThreadedSSAGraphExecutor(
size_t num_threads, bool use_event, size_t num_threads, bool use_event,
const std::vector<Scope *> &local_scopes, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
std::unique_ptr<SSAGraph> &&graph) std::unique_ptr<SSAGraph> &&graph, bool allow_op_delay)
: SSAGraphExecutor(std::move(graph)), : SSAGraphExecutor(std::move(graph)),
pool_(num_threads >= 2 ? new ::ThreadPool(num_threads) : nullptr), pool_(num_threads >= 2 ? new ::ThreadPool(num_threads) : nullptr),
local_scopes_(local_scopes), local_scopes_(local_scopes),
places_(places), places_(places),
fetch_ctxs_(places), fetch_ctxs_(places),
use_event_(use_event) {} use_event_(use_event),
running_ops_(0),
allow_op_delay_(allow_op_delay) {}
void ThreadedSSAGraphExecutor::RunDelayedOps(
const std::unordered_set<OpHandleBase *> &delayed_ops) {
for (auto op : delayed_ops) {
op->Run(use_event_);
}
}
FeedFetchList ThreadedSSAGraphExecutor::Run( FeedFetchList ThreadedSSAGraphExecutor::Run(
const std::vector<std::string> &fetch_tensors) { const std::vector<std::string> &fetch_tensors) {
std::unordered_map<OpHandleBase *, size_t> pending_ops; std::unordered_map<OpHandleBase *, size_t> pending_ops;
std::unordered_set<VarHandleBase *> pending_vars; std::unordered_set<VarHandleBase *> pending_vars;
BlockingQueue<VarHandleBase *> ready_vars; BlockingQueue<VarHandleBase *> ready_vars;
std::unordered_set<OpHandleBase *> ready_ops; std::unordered_set<OpHandleBase *> ready_ops;
// For ops (e.g. nccl_all_reduce) that need to coordinate multiple
// streams from multiple GPUs, it's faster to buffer them and schedule
// together since we currently cannot overlap computation and memcpy streams.
// Should revisit it if overlapping is available.
std::unordered_set<OpHandleBase *> delayed_ops;
std::unordered_set<OpHandleBase *> blocked_by_delayed_ops;
std::unordered_set<VarHandleBase *> delayed_vars;
auto InsertPendingVar = [&pending_vars, &ready_vars](VarHandleBase &var) { auto InsertPendingVar = [&pending_vars, &ready_vars](VarHandleBase &var) {
pending_vars.insert(&var); pending_vars.insert(&var);
...@@ -106,7 +120,14 @@ FeedFetchList ThreadedSSAGraphExecutor::Run( ...@@ -106,7 +120,14 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
auto run_all_ready_ops = [&] { auto run_all_ready_ops = [&] {
for (auto *op : ready_ops) { for (auto *op : ready_ops) {
RunOp(ready_vars, op); if (op->IsMultiDeviceTransfer() && allow_op_delay_) {
delayed_ops.insert(op);
delayed_vars.insert(op->outputs_.begin(), op->outputs_.end());
ready_vars.Extend(op->outputs_);
continue;
}
running_ops_++;
RunOp(&ready_vars, op);
} }
ready_ops.clear(); ready_ops.clear();
}; };
...@@ -118,13 +139,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run( ...@@ -118,13 +139,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
} }
// Step 3. Execution // Step 3. Execution
while (!pending_vars.empty()) { while (!pending_vars.empty() || !ready_ops.empty() || !delayed_ops.empty()) {
// 1. Run All Ready ops // 1. Run All Ready ops
run_all_ready_ops(); run_all_ready_ops();
// 2. Find ready variable // 2. Find ready variable
bool timeout; bool timeout;
auto cur_ready_vars = ready_vars.PopAll(1000, &timeout); auto cur_ready_vars = ready_vars.PopAll(1, &timeout);
if (timeout) { if (timeout) {
if (exception_) { if (exception_) {
...@@ -141,13 +162,29 @@ FeedFetchList ThreadedSSAGraphExecutor::Run( ...@@ -141,13 +162,29 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
auto &deps = pending_ops[op]; auto &deps = pending_ops[op];
--deps; --deps;
if (deps == 0) { if (deps == 0) {
ready_ops.insert(op); if (delayed_vars.find(ready_var) != delayed_vars.end()) {
blocked_by_delayed_ops.insert(op);
} else {
ready_ops.insert(op);
}
} }
} }
} }
// When there are no other ops to schedule, schedule buffered delayed
// ops and unblock other ops.
if (ready_ops.empty() && !delayed_ops.empty() && running_ops_ == 0) {
RunDelayedOps(delayed_ops);
delayed_ops.clear();
for (auto *op : blocked_by_delayed_ops) {
ready_ops.insert(op);
}
blocked_by_delayed_ops.clear();
}
// Keep loop until all vars are ready. // Keep loop until all vars are ready.
} }
PADDLE_ENFORCE(ready_ops.empty());
PADDLE_ENFORCE(delayed_ops.empty());
PADDLE_ENFORCE(blocked_by_delayed_ops.empty());
++computation_count_; ++computation_count_;
auto sync_computation = [&] { auto sync_computation = [&] {
...@@ -182,12 +219,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run( ...@@ -182,12 +219,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
} }
void ThreadedSSAGraphExecutor::RunOp( void ThreadedSSAGraphExecutor::RunOp(
BlockingQueue<VarHandleBase *> &ready_var_q, details::OpHandleBase *op) { BlockingQueue<VarHandleBase *> *ready_var_q, details::OpHandleBase *op) {
auto op_run = [&ready_var_q, op, this] { auto op_run = [ready_var_q, op, this] {
try { try {
VLOG(10) << op->Name() << " : " << op->DebugString(); VLOG(10) << op->Name() << " : " << op->DebugString();
op->Run(use_event_); op->Run(use_event_);
ready_var_q.Extend(op->outputs_); running_ops_--;
ready_var_q->Extend(op->outputs_);
} catch (platform::EnforceNotMet ex) { } catch (platform::EnforceNotMet ex) {
exception_.reset(new platform::EnforceNotMet(ex)); exception_.reset(new platform::EnforceNotMet(ex));
} catch (...) { } catch (...) {
......
paddle/fluid/framework/details/threaded_ssa_graph_executor.h
浏览文件 @ 1f6e0448
...@@ -14,7 +14,12 @@ ...@@ -14,7 +14,12 @@
#pragma once #pragma once
#include <chrono> #include <deque>
#include <string>
#include <unordered_set>
#include <utility>
#include <vector>
#include <functional> #include <functional>
#include "ThreadPool.h" // ThreadPool in thrird party #include "ThreadPool.h" // ThreadPool in thrird party
#include "paddle/fluid/framework/details/ssa_graph_executor.h" #include "paddle/fluid/framework/details/ssa_graph_executor.h"
...@@ -70,7 +75,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor { ...@@ -70,7 +75,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
ThreadedSSAGraphExecutor(size_t num_threads, bool use_event, ThreadedSSAGraphExecutor(size_t num_threads, bool use_event,
const std::vector<Scope *> &local_scopes, const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
std::unique_ptr<SSAGraph> &&graph); std::unique_ptr<SSAGraph> &&graph,
bool allow_op_delay);
// Run a SSAGraph by a thread pool // Run a SSAGraph by a thread pool
// Use topological sort algorithm // Use topological sort algorithm
...@@ -79,9 +85,11 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor { ...@@ -79,9 +85,11 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
~ThreadedSSAGraphExecutor() {} ~ThreadedSSAGraphExecutor() {}
private: private:
void RunOp(BlockingQueue<VarHandleBase *> &ready_var_q, void RunOp(BlockingQueue<VarHandleBase *> *ready_var_q,
details::OpHandleBase *op); details::OpHandleBase *op);
void RunDelayedOps(const std::unordered_set<OpHandleBase *> &delayed_ops);
private: private:
std::unique_ptr<::ThreadPool> pool_; std::unique_ptr<::ThreadPool> pool_;
std::vector<Scope *> local_scopes_; std::vector<Scope *> local_scopes_;
...@@ -89,6 +97,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor { ...@@ -89,6 +97,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
platform::DeviceContextPool fetch_ctxs_; platform::DeviceContextPool fetch_ctxs_;
const bool use_event_; const bool use_event_;
std::unique_ptr<platform::EnforceNotMet> exception_; std::unique_ptr<platform::EnforceNotMet> exception_;
std::atomic<int> running_ops_;
bool allow_op_delay_;
size_t computation_count_{0}; size_t computation_count_{0};
size_t max_async_computation{100}; size_t max_async_computation{100};
......
paddle/fluid/framework/details/var_handle.h
浏览文件 @ 1f6e0448
...@@ -22,7 +22,7 @@ ...@@ -22,7 +22,7 @@
namespace paddle { namespace paddle {
namespace framework { namespace framework {
namespace details { namespace details {
struct OpHandleBase; class OpHandleBase;
// VarHandleBase is the var node in the dependency graph. // VarHandleBase is the var node in the dependency graph.
// A variable can only be generated by a single operator. i.e. // A variable can only be generated by a single operator. i.e.
......
paddle/fluid/framework/parallel_executor.cc
浏览文件 @ 1f6e0448
...@@ -13,9 +13,10 @@ See the License for the specific language governing permissions and ...@@ -13,9 +13,10 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include "paddle/fluid/framework/parallel_executor.h" #include "paddle/fluid/framework/parallel_executor.h"
#include <string> #include "paddle/fluid/platform/profiler.h"
#include "ThreadPool.h" #include <string>
#include <vector>
#ifdef PADDLE_WITH_CUDA #ifdef PADDLE_WITH_CUDA
#include "paddle/fluid/platform/nccl_helper.h" #include "paddle/fluid/platform/nccl_helper.h"
...@@ -47,7 +48,7 @@ ParallelExecutor::ParallelExecutor( ...@@ -47,7 +48,7 @@ ParallelExecutor::ParallelExecutor(
const std::vector<platform::Place> &places, const std::vector<platform::Place> &places,
const std::unordered_set<std::string> &params, const std::unordered_set<std::string> &params,
const ProgramDesc &startup_program, const ProgramDesc &main_program, const ProgramDesc &startup_program, const ProgramDesc &main_program,
const std::string &loss_var_name, Scope *scope) const std::string &loss_var_name, Scope *scope, bool allow_op_delay)
: member_(new ParallelExecutorPrivate(places)) { : member_(new ParallelExecutorPrivate(places)) {
member_->global_scope_ = scope; member_->global_scope_ = scope;
...@@ -82,8 +83,8 @@ ParallelExecutor::ParallelExecutor( ...@@ -82,8 +83,8 @@ ParallelExecutor::ParallelExecutor(
auto graph = builder.Build(main_program); auto graph = builder.Build(main_program);
member_->executor_.reset(new details::ThreadedSSAGraphExecutor( member_->executor_.reset(new details::ThreadedSSAGraphExecutor(
num_threads, use_event, member_->local_scopes_, places, num_threads, use_event, member_->local_scopes_, places, std::move(graph),
std::move(graph))); allow_op_delay));
// Step 3. Create vars in each scope; // Step 3. Create vars in each scope;
for (auto *scope : member_->local_scopes_) { for (auto *scope : member_->local_scopes_) {
...@@ -151,6 +152,7 @@ void ParallelExecutor::BCastParamsToGPUs( ...@@ -151,6 +152,7 @@ void ParallelExecutor::BCastParamsToGPUs(
void ParallelExecutor::Run(const std::vector<std::string> &fetch_tensors, void ParallelExecutor::Run(const std::vector<std::string> &fetch_tensors,
const std::string &fetched_var_name) { const std::string &fetched_var_name) {
platform::RecordBlock b(0);
auto fetch_data = member_->executor_->Run(fetch_tensors); auto fetch_data = member_->executor_->Run(fetch_tensors);
*member_->global_scope_->Var(fetched_var_name)->GetMutable<FeedFetchList>() = *member_->global_scope_->Var(fetched_var_name)->GetMutable<FeedFetchList>() =
fetch_data; fetch_data;
......
paddle/fluid/framework/parallel_executor.h
浏览文件 @ 1f6e0448
...@@ -14,8 +14,9 @@ limitations under the License. */ ...@@ -14,8 +14,9 @@ limitations under the License. */
#pragma once #pragma once
#include <future> #include <string>
#include <unordered_set> #include <unordered_set>
#include <vector>
#include "paddle/fluid/framework/executor.h" #include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/op_info.h" #include "paddle/fluid/framework/op_info.h"
#include "paddle/fluid/framework/program_desc.h" #include "paddle/fluid/framework/program_desc.h"
...@@ -37,7 +38,8 @@ class ParallelExecutor { ...@@ -37,7 +38,8 @@ class ParallelExecutor {
const std::unordered_set<std::string>& params, const std::unordered_set<std::string>& params,
const ProgramDesc& startup_program, const ProgramDesc& startup_program,
const ProgramDesc& main_program, const ProgramDesc& main_program,
const std::string& loss_var_name, Scope* scope); const std::string& loss_var_name, Scope* scope,
bool allow_op_delay);
void Run(const std::vector<std::string>& fetch_tensors, void Run(const std::vector<std::string>& fetch_tensors,
const std::string& fetched_var_name = "fetched_var"); const std::string& fetched_var_name = "fetched_var");
......
paddle/fluid/framework/selected_rows.h
浏览文件 @ 1f6e0448
...@@ -10,6 +10,9 @@ See the License for the specific language governing permissions and ...@@ -10,6 +10,9 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#pragma once #pragma once
#include <vector>
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/tensor.h" #include "paddle/fluid/framework/tensor.h"
...@@ -52,7 +55,7 @@ class SelectedRows { ...@@ -52,7 +55,7 @@ class SelectedRows {
private: private:
// Notice: rows can be duplicate. We can have {0, 4, 7, 0, 5, 7, 9} here. // Notice: rows can be duplicate. We can have {0, 4, 7, 0, 5, 7, 9} here.
// SelectedRows are simplely concated when adding together. Until a // SelectedRows are simply concated when adding together. Until a
// SelectedRows add a Tensor, will the duplicate rows be handled. // SelectedRows add a Tensor, will the duplicate rows be handled.
Vector<int64_t> rows_; Vector<int64_t> rows_;
std::unique_ptr<Tensor> value_{nullptr}; std::unique_ptr<Tensor> value_{nullptr};
......
paddle/fluid/operators/detail/CMakeLists.txt
浏览文件 @ 1f6e0448
...@@ -2,7 +2,8 @@ if(WITH_DISTRIBUTE) ...@@ -2,7 +2,8 @@ if(WITH_DISTRIBUTE)
grpc_library(sendrecvop_grpc SRCS bytebuffer_stream.cc sendrecvop_utils.cc grpc_client.cc grpc_library(sendrecvop_grpc SRCS bytebuffer_stream.cc sendrecvop_utils.cc grpc_client.cc
grpc_server.cc variable_response.cc PROTO send_recv.proto DEPS lod_tensor selected_rows) grpc_server.cc variable_response.cc PROTO send_recv.proto DEPS lod_tensor selected_rows)
set(DISTRIBUTE_COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor") set(DISTRIBUTE_COMPILE_FLAGS "-Wno-non-virtual-dtor -Wno-error=non-virtual-dtor -Wno-error=delete-non-virtual-dtor")
set_source_files_properties(serde_test.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS}) set_source_files_properties(serde_test.cc grpc_server_test.cc PROPERTIES COMPILE_FLAGS ${DISTRIBUTE_COMPILE_FLAGS})
cc_test(serde_test SRCS serde_test.cc variable_response.cc DEPS grpc++_unsecure grpc_unsecure gpr cc_test(serde_test SRCS serde_test.cc variable_response.cc DEPS grpc++_unsecure grpc_unsecure gpr
cares zlib protobuf sendrecvop_grpc) cares zlib protobuf sendrecvop_grpc)
cc_test(grpc_server_test SRCS grpc_server_test.cc DEPS sendrecvop_grpc grpc++_unsecure grpc_unsecure gpr cares zlib protobuf)
endif() endif()
paddle/fluid/operators/detail/grpc_client.cc
浏览文件 @ 1f6e0448
...@@ -12,8 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. ...@@ -12,8 +12,10 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include "grpc_client.h" #include "paddle/fluid/operators/detail/grpc_client.h"
#include <sys/time.h>
#include <limits>
#include "paddle/fluid/framework/threadpool.h" #include "paddle/fluid/framework/threadpool.h"
namespace paddle { namespace paddle {
...@@ -52,7 +54,7 @@ bool RPCClient::AsyncSendVariable(const std::string& ep, ...@@ -52,7 +54,7 @@ bool RPCClient::AsyncSendVariable(const std::string& ep,
auto call = s->stub_g_.PrepareUnaryCall( auto call = s->stub_g_.PrepareUnaryCall(
s->context_.get(), "/sendrecv.SendRecvService/SendVariable", req, &cq_); s->context_.get(), "/sendrecv.SendRecvService/SendVariable", req, &cq_);
call->StartCall(); call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s); call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
}); });
req_count_++; req_count_++;
...@@ -70,8 +72,7 @@ void ProcGetResponse(const VarHandle& var_h, ...@@ -70,8 +72,7 @@ void ProcGetResponse(const VarHandle& var_h,
template <typename T> template <typename T>
void RequestToByteBuffer(const T& proto, ::grpc::ByteBuffer* result) { void RequestToByteBuffer(const T& proto, ::grpc::ByteBuffer* result) {
::grpc::Slice slice(proto.ByteSizeLong()); ::grpc::Slice slice(proto.ByteSizeLong());
proto.SerializeWithCachedSizesToArray( proto.SerializeWithCachedSizesToArray(const_cast<uint8_t*>(slice.begin()));
const_cast<uint8_t*>(reinterpret_cast<const uint8_t*>(slice.begin())));
::grpc::ByteBuffer tmp(&slice, 1); ::grpc::ByteBuffer tmp(&slice, 1);
result->Swap(&tmp); result->Swap(&tmp);
} }
...@@ -109,7 +110,7 @@ bool RPCClient::AsyncGetVariable(const std::string& ep, ...@@ -109,7 +110,7 @@ bool RPCClient::AsyncGetVariable(const std::string& ep,
auto call = s->stub_g_.PrepareUnaryCall( auto call = s->stub_g_.PrepareUnaryCall(
s->context_.get(), "/sendrecv.SendRecvService/GetVariable", buf, &cq_); s->context_.get(), "/sendrecv.SendRecvService/GetVariable", buf, &cq_);
call->StartCall(); call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s); call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
}); });
req_count_++; req_count_++;
...@@ -150,9 +151,10 @@ bool RPCClient::AsyncPrefetchVariable(const std::string& ep, ...@@ -150,9 +151,10 @@ bool RPCClient::AsyncPrefetchVariable(const std::string& ep,
s->response_call_back_ = ProcGetResponse; s->response_call_back_ = ProcGetResponse;
auto call = s->stub_g_.PrepareUnaryCall( auto call = s->stub_g_.PrepareUnaryCall(
s->context_.get(), "/sendrecv.SendRecvService/GetVariable", req, &cq_); s->context_.get(), "/sendrecv.SendRecvService/PrefetchVariable", req,
&cq_);
call->StartCall(); call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s); call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
}); });
req_count_++; req_count_++;
...@@ -168,7 +170,7 @@ void RPCClient::AsyncSendBatchBarrier(const std::string& ep, int64_t time_out) { ...@@ -168,7 +170,7 @@ void RPCClient::AsyncSendBatchBarrier(const std::string& ep, int64_t time_out) {
sendrecv::VariableMessage req; sendrecv::VariableMessage req;
req.set_varname(BATCH_BARRIER_MESSAGE); req.set_varname(BATCH_BARRIER_MESSAGE);
auto rpc = s->stub_->AsyncSendVariable(s->context_.get(), req, &cq_); auto rpc = s->stub_->AsyncSendVariable(s->context_.get(), req, &cq_);
rpc->Finish(&s->reply_, &s->status_, (void*)s); rpc->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
req_count_++; req_count_++;
} }
...@@ -180,7 +182,7 @@ void RPCClient::AsyncSendFetchBarrier(const std::string& ep, int64_t time_out) { ...@@ -180,7 +182,7 @@ void RPCClient::AsyncSendFetchBarrier(const std::string& ep, int64_t time_out) {
sendrecv::VariableMessage req; sendrecv::VariableMessage req;
req.set_varname(FETCH_BARRIER_MESSAGE); req.set_varname(FETCH_BARRIER_MESSAGE);
auto rpc = s->stub_->AsyncGetVariable(s->context_.get(), req, &cq_); auto rpc = s->stub_->AsyncGetVariable(s->context_.get(), req, &cq_);
rpc->Finish(&s->reply_, &s->status_, (void*)s); rpc->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
req_count_++; req_count_++;
} }
......
paddle/fluid/operators/detail/grpc_server.cc
浏览文件 @ 1f6e0448
...@@ -14,6 +14,9 @@ limitations under the License. */ ...@@ -14,6 +14,9 @@ limitations under the License. */
#include "paddle/fluid/operators/detail/grpc_server.h" #include "paddle/fluid/operators/detail/grpc_server.h"
#include <limits>
#include <string>
using ::grpc::ServerAsyncResponseWriter; using ::grpc::ServerAsyncResponseWriter;
namespace paddle { namespace paddle {
...@@ -128,6 +131,49 @@ class RequestGet final : public RequestBase { ...@@ -128,6 +131,49 @@ class RequestGet final : public RequestBase {
SimpleBlockQueue<MessageWithName>* queue_; SimpleBlockQueue<MessageWithName>* queue_;
}; };
class RequestPrefetch final : public RequestBase {
public:
explicit RequestPrefetch(GrpcService::AsyncService* service,
::grpc::ServerCompletionQueue* cq,
framework::Scope* scope,
const platform::DeviceContext* dev_ctx,
framework::Executor* executor,
framework::ProgramDesc* program, int blkid)
: RequestBase(service, cq, dev_ctx),
responder_(&ctx_),
scope_(scope),
executor_(executor),
program_(program),
blkid_(blkid) {
int method_id = static_cast<int>(detail::GrpcMethod::kPrefetchVariable);
service_->RequestAsyncUnary(method_id, &ctx_, &request_, &responder_, cq_,
cq_, this);
}
virtual ~RequestPrefetch() {}
virtual std::string GetReqName() { return request_.varname(); }
virtual void Process() {
// prefetch process...
::grpc::ByteBuffer reply;
// TODO(Yancey1989): execute the Block which containers prefetch ops
VLOG(3) << "RequestPrefetch Process in";
responder_.Finish(reply, ::grpc::Status::OK, this);
status_ = FINISH;
}
protected:
sendrecv::VariableMessage request_;
ServerAsyncResponseWriter<::grpc::ByteBuffer> responder_;
framework::Scope* scope_;
framework::Executor* executor_;
framework::ProgramDesc* program_;
int blkid_;
};
void AsyncGRPCServer::WaitClientGet(int count) { void AsyncGRPCServer::WaitClientGet(int count) {
int fetch_barriers = 0; int fetch_barriers = 0;
while (fetch_barriers < count) { while (fetch_barriers < count) {
...@@ -147,6 +193,7 @@ void AsyncGRPCServer::RunSyncUpdate() { ...@@ -147,6 +193,7 @@ void AsyncGRPCServer::RunSyncUpdate() {
cq_send_ = builder.AddCompletionQueue(); cq_send_ = builder.AddCompletionQueue();
cq_get_ = builder.AddCompletionQueue(); cq_get_ = builder.AddCompletionQueue();
cq_prefetch_ = builder.AddCompletionQueue();
server_ = builder.BuildAndStart(); server_ = builder.BuildAndStart();
LOG(INFO) << "Server listening on " << address_ << std::endl; LOG(INFO) << "Server listening on " << address_ << std::endl;
...@@ -155,6 +202,8 @@ void AsyncGRPCServer::RunSyncUpdate() { ...@@ -155,6 +202,8 @@ void AsyncGRPCServer::RunSyncUpdate() {
std::bind(&AsyncGRPCServer::TryToRegisterNewSendOne, this); std::bind(&AsyncGRPCServer::TryToRegisterNewSendOne, this);
std::function<void()> get_register = std::function<void()> get_register =
std::bind(&AsyncGRPCServer::TryToRegisterNewGetOne, this); std::bind(&AsyncGRPCServer::TryToRegisterNewGetOne, this);
std::function<void()> prefetch_register =
std::bind(&AsyncGRPCServer::TryToRegisterNewPrefetchOne, this);
t_send_.reset( t_send_.reset(
new std::thread(std::bind(&AsyncGRPCServer::HandleRequest, this, new std::thread(std::bind(&AsyncGRPCServer::HandleRequest, this,
...@@ -163,17 +212,21 @@ void AsyncGRPCServer::RunSyncUpdate() { ...@@ -163,17 +212,21 @@ void AsyncGRPCServer::RunSyncUpdate() {
t_get_.reset( t_get_.reset(
new std::thread(std::bind(&AsyncGRPCServer::HandleRequest, this, new std::thread(std::bind(&AsyncGRPCServer::HandleRequest, this,
cq_get_.get(), "cq_get", get_register))); cq_get_.get(), "cq_get", get_register)));
t_prefetch_.reset(new std::thread(
std::bind(&AsyncGRPCServer::HandleRequest, this, cq_prefetch_.get(),
"cq_prefetch", prefetch_register)));
// wait server // wait server
server_->Wait(); server_->Wait();
t_send_->join(); t_send_->join();
t_get_->join(); t_get_->join();
t_prefetch_->join();
} }
void AsyncGRPCServer::ShutdownQueue() { void AsyncGRPCServer::ShutdownQueue() {
std::unique_lock<std::mutex> lock(cq_mutex_); std::unique_lock<std::mutex> lock(cq_mutex_);
cq_send_->Shutdown(); cq_send_->Shutdown();
cq_get_->Shutdown(); cq_get_->Shutdown();
cq_prefetch_->Shutdown();
} }
// This URL explains why shutdown is complicate: // This URL explains why shutdown is complicate:
...@@ -186,6 +239,7 @@ void AsyncGRPCServer::ShutDown() { ...@@ -186,6 +239,7 @@ void AsyncGRPCServer::ShutDown() {
void AsyncGRPCServer::TryToRegisterNewSendOne() { void AsyncGRPCServer::TryToRegisterNewSendOne() {
std::unique_lock<std::mutex> lock(cq_mutex_); std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) { if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewSendOne";
return; return;
} }
RequestSend* send = new RequestSend(&service_, cq_send_.get(), scope_, RequestSend* send = new RequestSend(&service_, cq_send_.get(), scope_,
...@@ -196,6 +250,7 @@ void AsyncGRPCServer::TryToRegisterNewSendOne() { ...@@ -196,6 +250,7 @@ void AsyncGRPCServer::TryToRegisterNewSendOne() {
void AsyncGRPCServer::TryToRegisterNewGetOne() { void AsyncGRPCServer::TryToRegisterNewGetOne() {
std::unique_lock<std::mutex> lock(cq_mutex_); std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) { if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewGetOne";
return; return;
} }
RequestGet* get = new RequestGet(&service_, cq_get_.get(), scope_, dev_ctx_, RequestGet* get = new RequestGet(&service_, cq_get_.get(), scope_, dev_ctx_,
...@@ -203,6 +258,19 @@ void AsyncGRPCServer::TryToRegisterNewGetOne() { ...@@ -203,6 +258,19 @@ void AsyncGRPCServer::TryToRegisterNewGetOne() {
VLOG(4) << "Create RequestGet status:" << get->Status(); VLOG(4) << "Create RequestGet status:" << get->Status();
} }
void AsyncGRPCServer::TryToRegisterNewPrefetchOne() {
std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewPrefetchOne";
return;
}
RequestPrefetch* prefetch =
new RequestPrefetch(&service_, cq_prefetch_.get(), scope_, dev_ctx_,
executor_, program_, prefetch_blk_id_);
VLOG(4) << "Create RequestPrefetch status:" << prefetch->Status();
}
// FIXME(typhoonzero): change cq_name to enum. // FIXME(typhoonzero): change cq_name to enum.
void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq, void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq,
std::string cq_name, std::string cq_name,
...@@ -211,25 +279,28 @@ void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq, ...@@ -211,25 +279,28 @@ void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq,
void* tag = NULL; void* tag = NULL;
bool ok = false; bool ok = false;
while (true) { while (true) {
VLOG(3) << "HandleRequest for " << cq_name << " while in";
if (!cq->Next(&tag, &ok)) { if (!cq->Next(&tag, &ok)) {
LOG(INFO) << cq_name << " CompletionQueue shutdown!"; LOG(INFO) << cq_name << " CompletionQueue shutdown!";
break; break;
} }
VLOG(3) << "HandleRequest for " << cq_name << " while after Next";
PADDLE_ENFORCE(tag); PADDLE_ENFORCE(tag);
// FIXME(typhoonzero): de-couple the barriers with recv_op // FIXME(typhoonzero): de-couple the barriers with recv_op
if (!is_shut_down_ && cq_name == "cq_get") WaitCond(1); if (!is_shut_down_ && cq_name == "cq_get") WaitCond(1);
if (!is_shut_down_ && cq_name == "cq_send") WaitCond(0); if (!is_shut_down_ && cq_name == "cq_send") WaitCond(0);
RequestBase* base = (RequestBase*)tag; RequestBase* base = reinterpret_cast<RequestBase*>(tag);
// reference: // reference:
// https://github.com/tensorflow/tensorflow/issues/5596 // https://github.com/tensorflow/tensorflow/issues/5596
// https://groups.google.com/forum/#!topic/grpc-io/xftlRy-IQwM // https://groups.google.com/forum/#!topic/grpc-io/xftlRy-IQwM
// https://groups.google.com/forum/#!topic/grpc-io/ywATt88Ef_I // https://groups.google.com/forum/#!topic/grpc-io/ywATt88Ef_I
if (!ok) { if (!ok) {
LOG(WARNING) << cq_name << " recv no regular event:argument name" LOG(WARNING) << cq_name << " recv no regular event:argument name["
<< base->GetReqName(); << base->GetReqName() << "]";
TryToRegisterNewOne(); TryToRegisterNewOne();
delete base; delete base;
continue; continue;
......
paddle/fluid/operators/detail/grpc_server.h
浏览文件 @ 1f6e0448
...@@ -15,9 +15,12 @@ limitations under the License. */ ...@@ -15,9 +15,12 @@ limitations under the License. */
#pragma once #pragma once
#include <grpc++/grpc++.h> #include <grpc++/grpc++.h>
#include <thread> #include <string>
#include <utility>
#include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/program_desc.h"
#include "paddle/fluid/framework/scope.h" #include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/selected_rows.h" #include "paddle/fluid/framework/selected_rows.h"
#include "paddle/fluid/framework/var_type.h" #include "paddle/fluid/framework/var_type.h"
...@@ -53,6 +56,12 @@ class AsyncGRPCServer final { ...@@ -53,6 +56,12 @@ class AsyncGRPCServer final {
void SetDevCtx(const platform::DeviceContext *dev_ctx) { dev_ctx_ = dev_ctx; } void SetDevCtx(const platform::DeviceContext *dev_ctx) { dev_ctx_ = dev_ctx; }
void SetProgram(framework::ProgramDesc *program) { program_ = program; }
void SetPrefetchBlkdId(int blkid) { prefetch_blk_id_ = blkid; }
void SetExecutor(framework::Executor *executor) { executor_ = executor; }
const ReceivedMessage Get() { return this->var_recv_queue_.Pop(); } const ReceivedMessage Get() { return this->var_recv_queue_.Pop(); }
void Push(const std::string &msg_name) { void Push(const std::string &msg_name) {
...@@ -66,6 +75,7 @@ class AsyncGRPCServer final { ...@@ -66,6 +75,7 @@ class AsyncGRPCServer final {
std::function<void()> TryToRegisterNewOne); std::function<void()> TryToRegisterNewOne);
void TryToRegisterNewSendOne(); void TryToRegisterNewSendOne();
void TryToRegisterNewGetOne(); void TryToRegisterNewGetOne();
void TryToRegisterNewPrefetchOne();
void ShutdownQueue(); void ShutdownQueue();
private: private:
...@@ -73,6 +83,7 @@ class AsyncGRPCServer final { ...@@ -73,6 +83,7 @@ class AsyncGRPCServer final {
volatile bool is_shut_down_ = false; volatile bool is_shut_down_ = false;
std::unique_ptr<::grpc::ServerCompletionQueue> cq_send_; std::unique_ptr<::grpc::ServerCompletionQueue> cq_send_;
std::unique_ptr<::grpc::ServerCompletionQueue> cq_get_; std::unique_ptr<::grpc::ServerCompletionQueue> cq_get_;
std::unique_ptr<::grpc::ServerCompletionQueue> cq_prefetch_;
GrpcService::AsyncService service_; GrpcService::AsyncService service_;
std::unique_ptr<::grpc::Server> server_; std::unique_ptr<::grpc::Server> server_;
...@@ -83,6 +94,7 @@ class AsyncGRPCServer final { ...@@ -83,6 +94,7 @@ class AsyncGRPCServer final {
// received variable from RPC, operators fetch variable from this queue. // received variable from RPC, operators fetch variable from this queue.
SimpleBlockQueue<MessageWithName> var_get_queue_; SimpleBlockQueue<MessageWithName> var_get_queue_;
// client send variable to this queue.
ReceivedQueue var_recv_queue_; ReceivedQueue var_recv_queue_;
// condition of the sub program // condition of the sub program
...@@ -92,6 +104,11 @@ class AsyncGRPCServer final { ...@@ -92,6 +104,11 @@ class AsyncGRPCServer final {
std::unique_ptr<std::thread> t_send_; std::unique_ptr<std::thread> t_send_;
std::unique_ptr<std::thread> t_get_; std::unique_ptr<std::thread> t_get_;
std::unique_ptr<std::thread> t_prefetch_;
int prefetch_blk_id_;
framework::ProgramDesc *program_;
framework::Executor *executor_;
}; };
}; // namespace detail }; // namespace detail
......
paddle/fluid/operators/detail/grpc_server_test.cc 0 → 100644
浏览文件 @ 1f6e0448
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */
#include <unistd.h>
#include <string>
#include <thread>
#include "gtest/gtest.h"
#include "paddle/fluid/operators/detail/grpc_client.h"
#include "paddle/fluid/operators/detail/grpc_server.h"
namespace framework = paddle::framework;
namespace platform = paddle::platform;
namespace detail = paddle::operators::detail;
std::unique_ptr<detail::AsyncGRPCServer> rpc_service_;
void StartServer(const std::string& endpoint) {
rpc_service_.reset(new detail::AsyncGRPCServer(endpoint));
rpc_service_->RunSyncUpdate();
}
TEST(PREFETCH, CPU) {
// start up a server instance backend
// TODO(Yancey1989): Need to start a server with optimize blocks and
// prefetch blocks.
std::thread server_thread(StartServer, "127.0.0.1:8889");
framework::Scope scope;
platform::CPUPlace place;
platform::CPUDeviceContext ctx(place);
// create var on local scope
std::string in_var_name("in");
std::string out_var_name("out");
auto* in_var = scope.Var(in_var_name);
auto* in_tensor = in_var->GetMutable<framework::LoDTensor>();
in_tensor->Resize({10, 10});
VLOG(3) << "before mutable_data";
in_tensor->mutable_data<int>(place);
scope.Var(out_var_name);
VLOG(3) << "before fetch";
detail::RPCClient client;
client.AsyncPrefetchVariable("127.0.0.1:8889", ctx, scope, in_var_name,
out_var_name);
client.Wait();
rpc_service_->ShutDown();
server_thread.join();
rpc_service_.reset(nullptr);
}
paddle/fluid/operators/detail/grpc_service.h
浏览文件 @ 1f6e0448
...@@ -76,10 +76,11 @@ namespace detail { ...@@ -76,10 +76,11 @@ namespace detail {
enum class GrpcMethod { enum class GrpcMethod {
kSendVariable, kSendVariable,
kGetVariable, kGetVariable,
kPrefetchVariable,
}; };
static const int kGrpcNumMethods = static const int kGrpcNumMethods =
static_cast<int>(GrpcMethod::kGetVariable) + 1; static_cast<int>(GrpcMethod::kPrefetchVariable) + 1;
inline const char* GrpcMethodName(GrpcMethod id) { inline const char* GrpcMethodName(GrpcMethod id) {
switch (id) { switch (id) {
...@@ -87,6 +88,8 @@ inline const char* GrpcMethodName(GrpcMethod id) { ...@@ -87,6 +88,8 @@ inline const char* GrpcMethodName(GrpcMethod id) {
return "/sendrecv.SendRecvService/SendVariable"; return "/sendrecv.SendRecvService/SendVariable";
case GrpcMethod::kGetVariable: case GrpcMethod::kGetVariable:
return "/sendrecv.SendRecvService/GetVariable"; return "/sendrecv.SendRecvService/GetVariable";
case GrpcMethod::kPrefetchVariable:
return "/sendrecv.SendRecvService/PrefetchVariable";
} }
// Shouldn't be reached. // Shouldn't be reached.
...@@ -114,5 +117,5 @@ class GrpcService final { ...@@ -114,5 +117,5 @@ class GrpcService final {
}; };
} // namespace detail } // namespace detail
} // namespace operator } // namespace operators
} // namespace paddle } // namespace paddle
paddle/fluid/operators/detail/send_recv.proto
浏览文件 @ 1f6e0448
...@@ -21,6 +21,8 @@ service SendRecvService { ...@@ -21,6 +21,8 @@ service SendRecvService {
rpc SendVariable(VariableMessage) returns (VoidMessage) {} rpc SendVariable(VariableMessage) returns (VoidMessage) {}
// Argument VariableMessage for GetVariable should only contain varname. // Argument VariableMessage for GetVariable should only contain varname.
rpc GetVariable(VariableMessage) returns (VariableMessage) {} rpc GetVariable(VariableMessage) returns (VariableMessage) {}
// Prefetch variable by Ids
rpc PrefetchVariable(VariableMessage) returns (VariableMessage) {}
} }
// VariableMessage is serialized paddle variable message. // VariableMessage is serialized paddle variable message.
......
paddle/fluid/operators/listen_and_serv_op.cc
浏览文件 @ 1f6e0448
...@@ -13,22 +13,13 @@ See the License for the specific language governing permissions and ...@@ -13,22 +13,13 @@ See the License for the specific language governing permissions and
limitations under the License. */ limitations under the License. */
#include <stdint.h> #include <stdint.h>
#include <sys/stat.h>
#include <ostream> #include <ostream>
#include <thread>
#include <unistd.h>
#include "paddle/fluid/framework/executor.h" #include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/framework.pb.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/op_registry.h" #include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/proto_desc.h"
#include "paddle/fluid/framework/threadpool.h" #include "paddle/fluid/framework/threadpool.h"
#include "paddle/fluid/operators/detail/grpc_server.h" #include "paddle/fluid/operators/detail/grpc_server.h"
#include "paddle/fluid/operators/detail/sendrecvop_utils.h"
#include "paddle/fluid/operators/detail/simple_block_queue.h"
#include "paddle/fluid/string/printf.h"
namespace paddle { namespace paddle {
namespace operators { namespace operators {
...@@ -121,6 +112,11 @@ class ListenAndServOp : public framework::OperatorBase { ...@@ -121,6 +112,11 @@ class ListenAndServOp : public framework::OperatorBase {
prepared.begin(), prepared.begin(),
std::shared_ptr<framework::ExecutorPrepareContext>(nullptr)); std::shared_ptr<framework::ExecutorPrepareContext>(nullptr));
// TODO(qiao) set proper fields for table lookup and update
rpc_service_->SetExecutor(&executor);
rpc_service_->SetPrefetchBlkdId(0);
rpc_service_->SetProgram(program);
// TODO(typhoonzero): change this to a while_op for every cluster-batch. // TODO(typhoonzero): change this to a while_op for every cluster-batch.
bool exit_flag = false; bool exit_flag = false;
// Record received sparse variables, so that // Record received sparse variables, so that
...@@ -184,7 +180,8 @@ class ListenAndServOp : public framework::OperatorBase { ...@@ -184,7 +180,8 @@ class ListenAndServOp : public framework::OperatorBase {
ParallelExecuteBlocks(parallel_blkids, &executor, prepared, program, ParallelExecuteBlocks(parallel_blkids, &executor, prepared, program,
&recv_scope); &recv_scope);
VLOG(2) << "run all blocks spent (ms) " << detail::GetTimestamp() - ts; VLOG(3) << "run all blocks spent " << detail::GetTimestamp() - ts
<< "(ms)";
// Reset the received sparse variables, the sum operator would not // Reset the received sparse variables, the sum operator would not
// sum the input sparse variables which rows is empty at the next // sum the input sparse variables which rows is empty at the next
......
paddle/fluid/operators/lookup_table_op.cc
浏览文件 @ 1f6e0448
...@@ -18,6 +18,22 @@ limitations under the License. */ ...@@ -18,6 +18,22 @@ limitations under the License. */
namespace paddle { namespace paddle {
namespace operators { namespace operators {
static inline framework::OpKernelType ExpectedKernelType(
const framework::ExecutionContext& ctx) {
auto* table_var = ctx.InputVar("W");
if (table_var->IsType<LoDTensor>()) {
return framework::OpKernelType(
framework::ToDataType(table_var->Get<LoDTensor>().type()),
ctx.device_context());
} else if (table_var->IsType<SelectedRows>()) {
return framework::OpKernelType(
framework::ToDataType(table_var->Get<SelectedRows>().value().type()),
ctx.device_context());
} else {
PADDLE_THROW("W should be LoDTensor or SelectedRows");
}
}
class LookupTableOp : public framework::OperatorWithKernel { class LookupTableOp : public framework::OperatorWithKernel {
public: public:
using framework::OperatorWithKernel::OperatorWithKernel; using framework::OperatorWithKernel::OperatorWithKernel;
...@@ -51,9 +67,7 @@ class LookupTableOp : public framework::OperatorWithKernel { ...@@ -51,9 +67,7 @@ class LookupTableOp : public framework::OperatorWithKernel {
protected: protected:
framework::OpKernelType GetExpectedKernelType( framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext& ctx) const override { const framework::ExecutionContext& ctx) const override {
return framework::OpKernelType( return ExpectedKernelType(ctx);
framework::ToDataType(ctx.Input<LoDTensor>("W")->type()),
ctx.device_context());
} }
}; };
...@@ -84,7 +98,7 @@ class LookupTableOpMaker : public framework::OpProtoAndCheckerMaker { ...@@ -84,7 +98,7 @@ class LookupTableOpMaker : public framework::OpProtoAndCheckerMaker {
"If the value is -1, it makes no effect to lookup. " "If the value is -1, it makes no effect to lookup. "
"Otherwise the given value indicates padding the output " "Otherwise the given value indicates padding the output "
"with zeros whenever lookup encounters it in Ids.") "with zeros whenever lookup encounters it in Ids.")
.SetDefault(-1); .SetDefault(kNoPadding);
AddComment(R"DOC( AddComment(R"DOC(
Lookup Table Operator. Lookup Table Operator.
...@@ -124,9 +138,7 @@ class LookupTableOpGrad : public framework::OperatorWithKernel { ...@@ -124,9 +138,7 @@ class LookupTableOpGrad : public framework::OperatorWithKernel {
protected: protected:
framework::OpKernelType GetExpectedKernelType( framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext& ctx) const override { const framework::ExecutionContext& ctx) const override {
return framework::OpKernelType( return ExpectedKernelType(ctx);
framework::ToDataType(ctx.Input<LoDTensor>("W")->type()),
ctx.device_context());
} }
}; };
......
paddle/fluid/operators/lookup_table_op.h
浏览文件 @ 1f6e0448
...@@ -14,6 +14,9 @@ limitations under the License. */ ...@@ -14,6 +14,9 @@ limitations under the License. */
#pragma once #pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/eigen.h" #include "paddle/fluid/framework/eigen.h"
#include "paddle/fluid/framework/lod_tensor.h" #include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/op_registry.h" #include "paddle/fluid/framework/op_registry.h"
...@@ -25,16 +28,37 @@ namespace operators { ...@@ -25,16 +28,37 @@ namespace operators {
using Tensor = framework::Tensor; using Tensor = framework::Tensor;
using LoDTensor = framework::LoDTensor; using LoDTensor = framework::LoDTensor;
using SelectedRows = framework::SelectedRows; using SelectedRows = framework::SelectedRows;
using DDim = framework::DDim;
static constexpr int64_t kNoPadding = -1;
inline size_t getIndex(const std::vector<int64_t> &rows, int64_t value) {
auto it = std::find(rows.begin(), rows.end(), value);
PADDLE_ENFORCE(it != rows.end(), "id should be in rows");
return static_cast<size_t>(std::distance(rows.begin(), it));
}
template <typename T> template <typename T>
class LookupTableKernel : public framework::OpKernel<T> { class LookupTableKernel : public framework::OpKernel<T> {
public: public:
void Compute(const framework::ExecutionContext& context) const override { void Compute(const framework::ExecutionContext &context) const override {
auto* table_t = context.Input<LoDTensor>("W"); auto *table_var = context.InputVar("W");
auto* ids_var = context.InputVar("Ids"); auto *ids_var = context.InputVar("Ids");
Tensor* output_t = context.Output<Tensor>("Out"); Tensor *output_t = context.Output<Tensor>("Out");
int64_t padding_idx = context.Attr<int64_t>("padding_idx");
DDim table_dim;
int64_t* ids; if (table_var->IsType<LoDTensor>()) {
table_dim = context.Input<LoDTensor>("W")->dims();
} else if (table_var->IsType<SelectedRows>()) {
auto *table_t = context.Input<SelectedRows>("W");
table_dim = table_t->value().dims();
} else {
PADDLE_THROW("table only support LoDTensor and SelectedRows");
}
int64_t *ids;
int64_t ids_numel; int64_t ids_numel;
// The type of Ids(Input) is SelectedRows or LoDTensor, when Ids's type // The type of Ids(Input) is SelectedRows or LoDTensor, when Ids's type
...@@ -42,39 +66,50 @@ class LookupTableKernel : public framework::OpKernel<T> { ...@@ -42,39 +66,50 @@ class LookupTableKernel : public framework::OpKernel<T> {
// when Ids's type is SelectedRows, the rows of Ids contains the // when Ids's type is SelectedRows, the rows of Ids contains the
// ids to be looked up in W. // ids to be looked up in W.
if (ids_var->IsType<LoDTensor>()) { if (ids_var->IsType<LoDTensor>()) {
auto* ids_t = context.Input<LoDTensor>("Ids"); auto *ids_t = context.Input<LoDTensor>("Ids");
ids = const_cast<int64_t*>(ids_t->data<int64_t>()); ids = const_cast<int64_t *>(ids_t->data<int64_t>());
ids_numel = ids_t->numel(); ids_numel = ids_t->numel();
} else if (ids_var->IsType<SelectedRows>()) { } else if (ids_var->IsType<SelectedRows>()) {
auto* ids_t = context.Input<SelectedRows>("Ids"); auto *ids_t = context.Input<SelectedRows>("Ids");
ids = const_cast<int64_t*>(ids_t->rows().data()); ids = const_cast<int64_t *>(ids_t->rows().data());
ids_numel = ids_t->rows().size(); ids_numel = ids_t->rows().size();
output_t->Resize({ids_numel, table_t->dims()[1]}); output_t->Resize({ids_numel, table_dim[1]});
} else { } else {
PADDLE_THROW("Unsupported Variable Type of Ids"); PADDLE_THROW("Unsupported Variable Type of Ids");
} }
int64_t padding_idx = context.Attr<int64_t>("padding_idx"); if (table_var->IsType<LoDTensor>()) {
auto *table_t = context.Input<LoDTensor>("W");
int64_t row_number = table_t->dims()[0];
int64_t row_width = table_t->dims()[1];
int N = table_t->dims()[0]; auto *table = table_t->data<T>();
int D = table_t->dims()[1]; auto *output = output_t->mutable_data<T>(context.GetPlace());
auto* table = table_t->data<T>();
auto* output = output_t->mutable_data<T>(context.GetPlace());
if (padding_idx == -1) {
for (int64_t i = 0; i < ids_numel; ++i) { for (int64_t i = 0; i < ids_numel; ++i) {
PADDLE_ENFORCE_LT(ids[i], N); if (padding_idx != kNoPadding && ids[i] == padding_idx) {
PADDLE_ENFORCE_GE(ids[i], 0); memset(output + i * row_width, 0, row_width * sizeof(T));
memcpy(output + i * D, table + ids[i] * D, D * sizeof(T)); } else {
PADDLE_ENFORCE_LT(ids[i], row_number);
PADDLE_ENFORCE_GE(ids[i], 0);
memcpy(output + i * row_width, table + ids[i] * row_width,
row_width * sizeof(T));
}
} }
} else { } else if (table_var->IsType<SelectedRows>()) {
const auto &table_t = table_var->Get<SelectedRows>();
int64_t row_width = table_t.value().dims()[1];
const auto *table = table_t.value().data<T>();
auto *output = output_t->mutable_data<T>(context.GetPlace());
for (int64_t i = 0; i < ids_numel; ++i) { for (int64_t i = 0; i < ids_numel; ++i) {
if (ids[i] == padding_idx) { if (padding_idx != kNoPadding && ids[i] == padding_idx) {
memset(output + i * D, 0, D * sizeof(T)); memset(output + i * row_width, 0, row_width * sizeof(T));
} else { } else {
PADDLE_ENFORCE_LT(ids[i], N);
PADDLE_ENFORCE_GE(ids[i], 0); PADDLE_ENFORCE_GE(ids[i], 0);
memcpy(output + i * D, table + ids[i] * D, D * sizeof(T)); auto id_index = getIndex(table_t.rows(), ids[i]);
memcpy(output + i * row_width, table + id_index * row_width,
row_width * sizeof(T));
} }
} }
} }
...@@ -84,17 +119,27 @@ class LookupTableKernel : public framework::OpKernel<T> { ...@@ -84,17 +119,27 @@ class LookupTableKernel : public framework::OpKernel<T> {
template <typename T> template <typename T>
class LookupTableGradKernel : public framework::OpKernel<T> { class LookupTableGradKernel : public framework::OpKernel<T> {
public: public:
void Compute(const framework::ExecutionContext& context) const override { void Compute(const framework::ExecutionContext &context) const override {
auto *table_var = context.InputVar("W");
DDim table_dim;
if (table_var->IsType<LoDTensor>()) {
table_dim = context.Input<LoDTensor>("W")->dims();
} else if (table_var->IsType<SelectedRows>()) {
auto *table_t = context.Input<SelectedRows>("W");
table_dim = table_t->value().dims();
} else {
PADDLE_THROW("table only support LoDTensor and SelectedRows");
}
bool is_sparse = context.Attr<bool>("is_sparse"); bool is_sparse = context.Attr<bool>("is_sparse");
// Since paddings are not trainable and fixed in forward, the gradient of // Since paddings are not trainable and fixed in forward, the gradient of
// paddings makes no sense and we don't deal with it in backward. // paddings makes no sense and we don't deal with it in backward.
if (is_sparse) { if (is_sparse) {
auto* ids = context.Input<LoDTensor>("Ids"); auto *ids = context.Input<LoDTensor>("Ids");
auto* table = context.Input<LoDTensor>("W"); auto *d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
auto* d_output = context.Input<LoDTensor>(framework::GradVarName("Out")); auto *d_table = context.Output<SelectedRows>(framework::GradVarName("W"));
auto* d_table = context.Output<SelectedRows>(framework::GradVarName("W"));
auto* ids_data = ids->data<int64_t>(); auto *ids_data = ids->data<int64_t>();
auto ids_dim = ids->dims(); auto ids_dim = ids->dims();
framework::Vector<int64_t> new_rows; framework::Vector<int64_t> new_rows;
...@@ -104,31 +149,30 @@ class LookupTableGradKernel : public framework::OpKernel<T> { ...@@ -104,31 +149,30 @@ class LookupTableGradKernel : public framework::OpKernel<T> {
} }
d_table->set_rows(new_rows); d_table->set_rows(new_rows);
auto* d_table_value = d_table->mutable_value(); auto *d_table_value = d_table->mutable_value();
d_table_value->Resize({ids_dim[0], table->dims()[1]}); d_table_value->Resize({ids_dim[0], table_dim[1]});
d_table_value->mutable_data<T>(context.GetPlace()); d_table_value->mutable_data<T>(context.GetPlace());
d_table->set_height(table->dims()[0]); d_table->set_height(table_dim[0]);
auto* d_output_data = d_output->data<T>(); auto *d_output_data = d_output->data<T>();
auto* d_table_data = d_table_value->data<T>(); auto *d_table_data = d_table_value->data<T>();
PADDLE_ENFORCE_EQ(d_table_value->dims(), d_output->dims()); PADDLE_ENFORCE_EQ(d_table_value->dims(), d_output->dims());
memcpy(d_table_data, d_output_data, sizeof(T) * d_output->numel()); memcpy(d_table_data, d_output_data, sizeof(T) * d_output->numel());
} else { } else {
auto* ids = context.Input<LoDTensor>("Ids"); auto *ids = context.Input<LoDTensor>("Ids");
auto* d_output = context.Input<LoDTensor>(framework::GradVarName("Out")); auto *d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
auto* d_table = context.Output<LoDTensor>(framework::GradVarName("W")); auto *d_table = context.Output<LoDTensor>(framework::GradVarName("W"));
auto* table = context.Input<LoDTensor>("W");
auto* ids_data = ids->data<int64_t>(); auto *ids_data = ids->data<int64_t>();
auto ids_dim = ids->dims(); auto ids_dim = ids->dims();
int N = table->dims()[0]; int N = table_dim[0];
int D = d_output->dims()[1]; int D = d_output->dims()[1];
auto* d_output_data = d_output->data<T>(); auto *d_output_data = d_output->data<T>();
auto* d_table_data = d_table->mutable_data<T>(context.GetPlace()); auto *d_table_data = d_table->mutable_data<T>(context.GetPlace());
memset(d_table_data, 0, d_table->numel() * sizeof(T)); memset(d_table_data, 0, d_table->numel() * sizeof(T));
......
paddle/fluid/operators/nccl_op_test.cu.cc
浏览文件 @ 1f6e0448
...@@ -137,6 +137,8 @@ class NCCLTester : public ::testing::Test { ...@@ -137,6 +137,8 @@ class NCCLTester : public ::testing::Test {
TEST_F(NCCLTester, ncclInitOp) {} TEST_F(NCCLTester, ncclInitOp) {}
// ncclAllReduceOp with desc // ncclAllReduceOp with desc
// TODO(helin): https://github.com/PaddlePaddle/Paddle/issues/9367
/*
TEST_F(NCCLTester, ncclAllReduceOp) { TEST_F(NCCLTester, ncclAllReduceOp) {
std::unique_ptr<f::OpDesc> op2(new f::OpDesc); std::unique_ptr<f::OpDesc> op2(new f::OpDesc);
op2->SetType("ncclAllReduce"); op2->SetType("ncclAllReduce");
...@@ -184,6 +186,7 @@ TEST_F(NCCLTester, ncclAllReduceOp) { ...@@ -184,6 +186,7 @@ TEST_F(NCCLTester, ncclAllReduceOp) {
} }
} }
} }
*/
// ncclReduceOp with desc // ncclReduceOp with desc
TEST_F(NCCLTester, ncclReduceOp) { TEST_F(NCCLTester, ncclReduceOp) {
...@@ -236,6 +239,8 @@ TEST_F(NCCLTester, ncclReduceOp) { ...@@ -236,6 +239,8 @@ TEST_F(NCCLTester, ncclReduceOp) {
} }
// ncclBcastOp with desc // ncclBcastOp with desc
// TODO(helin): https://github.com/PaddlePaddle/Paddle/issues/9540
/*
TEST_F(NCCLTester, ncclBcastOp) { TEST_F(NCCLTester, ncclBcastOp) {
std::unique_ptr<f::OpDesc> op2(new f::OpDesc); std::unique_ptr<f::OpDesc> op2(new f::OpDesc);
const int kRoot = 0; const int kRoot = 0;
...@@ -281,3 +286,4 @@ TEST_F(NCCLTester, ncclBcastOp) { ...@@ -281,3 +286,4 @@ TEST_F(NCCLTester, ncclBcastOp) {
ASSERT_NEAR(ct[j], result, 1e-5); ASSERT_NEAR(ct[j], result, 1e-5);
} }
} }
*/
paddle/fluid/operators/reader/create_double_buffer_reader_op.cc
浏览文件 @ 1f6e0448
...@@ -20,12 +20,29 @@ namespace paddle { ...@@ -20,12 +20,29 @@ namespace paddle {
namespace operators { namespace operators {
namespace reader { namespace reader {
static constexpr size_t kDoubleBufferSize = 2; // 'Double buffer' means we shall maintain two batches of input data at the same
// time. So the kCacheSize shoul be at least 2.
static constexpr size_t kCacheSize = 2;
// There will be two bacthes out of the channel during training:
// 1. the one waiting to be sent to the channel
// 2. the one just be received from the channel, which is also being used by
// subsequent operators.
// So the channel size should be kChacheSize - 2
static constexpr size_t kChannelSize = 0; // kCacheSize - 2
class DoubleBufferReader : public framework::DecoratedReader { class DoubleBufferReader : public framework::DecoratedReader {
public: public:
struct Item { struct Item {
Item() : ctx_(nullptr) {} Item() : ctx_(nullptr) {}
Item(Item&& b) {
payloads_ = std::move(b.payloads_);
ctx_ = std::move(b.ctx_);
}
Item& operator=(Item&& b) {
payloads_ = std::move(b.payloads_);
ctx_ = std::move(b.ctx_);
return *this;
}
std::vector<framework::LoDTensor> payloads_; std::vector<framework::LoDTensor> payloads_;
platform::DeviceContext* ctx_; platform::DeviceContext* ctx_;
...@@ -34,42 +51,44 @@ class DoubleBufferReader : public framework::DecoratedReader { ...@@ -34,42 +51,44 @@ class DoubleBufferReader : public framework::DecoratedReader {
explicit DoubleBufferReader( explicit DoubleBufferReader(
ReaderBase* reader, platform::Place target_place = platform::CPUPlace()) ReaderBase* reader, platform::Place target_place = platform::CPUPlace())
: DecoratedReader(reader), place_(target_place) { : DecoratedReader(reader), place_(target_place) {
for (size_t i = 0; i < kDoubleBufferSize; ++i) {
if (platform::is_gpu_place(place_)) {
#ifdef PADDLE_WITH_CUDA #ifdef PADDLE_WITH_CUDA
for (size_t i = 0; i < kCacheSize; ++i) {
if (platform::is_gpu_place(place_)) {
ctxs_.emplace_back(new platform::CUDADeviceContext( ctxs_.emplace_back(new platform::CUDADeviceContext(
boost::get<platform::CUDAPlace>(place_))); boost::get<platform::CUDAPlace>(place_)));
#endif
} }
} }
#endif
start_thread(); StartPrefetcher();
}
void start_thread() {
buffer_ = framework::MakeChannel<Item>(kDoubleBufferSize);
prefetcher_ = std::thread([this] { PrefetchThreadFunc(); });
} }
bool HasNext() const override;
void ReadNext(std::vector<framework::LoDTensor>* out) override; void ReadNext(std::vector<framework::LoDTensor>* out) override;
void ReInit() override; void ReInit() override;
~DoubleBufferReader() { ~DoubleBufferReader() { EndPrefetcher(); }
buffer_->Close();
prefetcher_.join(); private:
delete buffer_; void StartPrefetcher() {
channel_ = framework::MakeChannel<Item>(kChannelSize);
prefetcher_ = std::thread([this] { PrefetchThreadFunc(); });
} }
bool HasNext() const override; void EndPrefetcher() {
channel_->Close();
if (prefetcher_.joinable()) {
prefetcher_.join();
}
delete channel_;
channel_ = nullptr;
}
private:
void PrefetchThreadFunc(); void PrefetchThreadFunc();
std::thread prefetcher_; std::thread prefetcher_;
framework::Channel<Item>* buffer_; framework::Channel<Item>* channel_;
platform::Place place_; platform::Place place_;
std::vector<std::unique_ptr<platform::DeviceContext>> ctxs_; std::vector<std::unique_ptr<platform::DeviceContext>> ctxs_;
mutable Item local_buffer_;
}; };
class CreateDoubleBufferReaderOp : public framework::OperatorBase { class CreateDoubleBufferReaderOp : public framework::OperatorBase {
...@@ -123,70 +142,70 @@ class CreateDoubleBufferReaderOpMaker : public DecoratedReaderMakerBase { ...@@ -123,70 +142,70 @@ class CreateDoubleBufferReaderOpMaker : public DecoratedReaderMakerBase {
} }
}; };
bool DoubleBufferReader::HasNext() const {
while (!channel_->IsClosed() && !channel_->CanReceive()) {
}
return channel_->CanReceive();
}
void DoubleBufferReader::ReadNext(std::vector<framework::LoDTensor>* out) { void DoubleBufferReader::ReadNext(std::vector<framework::LoDTensor>* out) {
if (!HasNext()) { if (!HasNext()) {
PADDLE_THROW("There is no next data!"); PADDLE_THROW("There is no next data!");
} }
if (local_buffer_.payloads_.empty()) { Item batch;
buffer_->Receive(&local_buffer_); channel_->Receive(&batch);
} *out = batch.payloads_;
*out = local_buffer_.payloads_; if (batch.ctx_) {
local_buffer_.payloads_.clear(); batch.ctx_->Wait();
if (local_buffer_.ctx_) {
local_buffer_.ctx_->Wait();
} }
} }
void DoubleBufferReader::ReInit() { void DoubleBufferReader::ReInit() {
reader_->ReInit(); reader_->ReInit();
buffer_->Close(); EndPrefetcher();
prefetcher_.join(); StartPrefetcher();
delete buffer_;
start_thread();
} }
void DoubleBufferReader::PrefetchThreadFunc() { void DoubleBufferReader::PrefetchThreadFunc() {
VLOG(5) << "A new prefetch thread starts."; VLOG(5) << "A new prefetch thread starts.";
size_t gpu_ctx_offset = 0; std::vector<std::vector<framework::LoDTensor>> cpu_tensor_cache(kCacheSize);
std::vector<std::vector<framework::LoDTensor>> gpu_tensor_cache(kCacheSize);
size_t cached_tensor_id = 0;
while (reader_->HasNext()) { while (reader_->HasNext()) {
Item batch; Item batch;
reader_->ReadNext(&batch.payloads_); auto& cpu_batch = cpu_tensor_cache[cached_tensor_id];
reader_->ReadNext(&cpu_batch);
if (platform::is_gpu_place(place_)) { if (platform::is_gpu_place(place_)) {
std::vector<framework::LoDTensor> gpu_batch; auto& gpu_batch = gpu_tensor_cache[cached_tensor_id];
auto& gpu_ctx = this->ctxs_[gpu_ctx_offset++]; auto* gpu_ctx = ctxs_[cached_tensor_id].get();
gpu_ctx_offset %= this->ctxs_.size(); gpu_batch.resize(cpu_batch.size());
gpu_batch.resize(batch.payloads_.size()); for (size_t i = 0; i < cpu_batch.size(); ++i) {
for (size_t i = 0; i < batch.payloads_.size(); ++i) { framework::TensorCopy(cpu_batch[i], place_, *gpu_ctx, &gpu_batch[i]);
framework::TensorCopy(batch.payloads_[i], place_, *gpu_ctx, gpu_batch[i].set_lod(cpu_batch[i].lod());
&gpu_batch[i]);
gpu_batch[i].set_lod(batch.payloads_[i].lod());
} }
batch.ctx_ = gpu_ctx.get(); batch.payloads_ = gpu_batch;
std::swap(gpu_batch, batch.payloads_); batch.ctx_ = gpu_ctx;
} else {
// CPUPlace
batch.payloads_ = cpu_batch;
} }
++cached_tensor_id;
cached_tensor_id %= kCacheSize;
try { try {
buffer_->Send(&batch); channel_->Send(&batch);
} catch (paddle::platform::EnforceNotMet e) { } catch (paddle::platform::EnforceNotMet e) {
VLOG(5) << "WARNING: The double buffer channel has been closed. The " VLOG(5) << "WARNING: The double buffer channel has been closed. The "
"prefetch thread will terminate."; "prefetch thread will terminate.";
break; break;
} }
} }
buffer_->Close(); channel_->Close();
VLOG(5) << "Prefetch thread terminates."; VLOG(5) << "Prefetch thread terminates.";
} }
bool DoubleBufferReader::HasNext() const {
if (local_buffer_.payloads_.empty()) {
bool ok = buffer_->Receive(&local_buffer_);
return ok;
} else {
return true;
}
}
} // namespace reader } // namespace reader
} // namespace operators } // namespace operators
} // namespace paddle } // namespace paddle
......
paddle/fluid/operators/reshape_op.cc
浏览文件 @ 1f6e0448
...@@ -17,90 +17,66 @@ limitations under the License. */ ...@@ -17,90 +17,66 @@ limitations under the License. */
namespace paddle { namespace paddle {
namespace operators { namespace operators {
class ReshapeOp : public framework::OperatorWithKernel {
public:
ReshapeOp(const std::string &type, const framework::VariableNameMap &inputs,
const framework::VariableNameMap &outputs,
const framework::AttributeMap &attrs)
: OperatorWithKernel(type, inputs, outputs, attrs) {}
void InferShape(framework::InferShapeContext *ctx) const override {
// input check
PADDLE_ENFORCE(ctx->HasInput("X"),
"Input(X) of ReshapeOp should not be null.");
PADDLE_ENFORCE(ctx->HasOutput("Out"),
"Output(Out) of ReshapeOp should not be null.");
auto shape = ctx->Attrs().Get<std::vector<int>>("shape");
PADDLE_ENFORCE(shape.size() > 0, "Attr(shape) shouldn't be empty.");
auto x_dims = ctx->GetInputDim("X");
std::vector<size_t> neg_dims_idx;
// set some dimension to -1 if it is unknown
const int unknown_size = -1;
for (size_t i = 0; i < shape.size(); ++i) {
PADDLE_ENFORCE(shape[i] > 0 || shape[i] == unknown_size,
"Each dimension of Attr(shape) must be positive or %d.",
unknown_size);
if (shape[i] == unknown_size) {
neg_dims_idx.push_back(i);
PADDLE_ENFORCE(neg_dims_idx.size() <= 1,
"Only one dimension of Attr(shape) can be unknown.");
}
}
int64_t capacity =
std::accumulate(shape.begin(), shape.end(), 1, std::multiplies<int>());
int64_t in_size = framework::product(x_dims);
if (neg_dims_idx.size() == 1) {
// dim infer
shape[neg_dims_idx[0]] = in_size / (-capacity);
// recalculate capacity
capacity = shape[neg_dims_idx[0]] * (-capacity);
}
// capacity check
PADDLE_ENFORCE(capacity == in_size,
"The size of Input(X) mismatches with Attr(shape).");
// resize output
std::vector<int64_t> shape_int64(shape.size(), 0);
std::transform(shape.begin(), shape.end(), shape_int64.begin(),
[](int a) { return static_cast<int64_t>(a); });
auto out_dims = framework::make_ddim(shape_int64);
ctx->SetOutputDim("Out", out_dims);
if (shape[0] == x_dims[0]) {
// Only pass LoD when the first dimension is equal between
// output and input.
ctx->ShareLoD("X", /*->*/ "Out");
}
}
};
class ReshapeOpMaker : public framework::OpProtoAndCheckerMaker { class ReshapeOpMaker : public framework::OpProtoAndCheckerMaker {
public: public:
ReshapeOpMaker(OpProto *proto, OpAttrChecker *op_checker) ReshapeOpMaker(OpProto *proto, OpAttrChecker *op_checker)
: OpProtoAndCheckerMaker(proto, op_checker) { : OpProtoAndCheckerMaker(proto, op_checker) {
AddInput("X", "The input tensor of reshape operator."); AddInput("X", "(Tensor). The input tensor of reshape operator.");
AddOutput("Out", "The output tensor of reshape operator."); AddInput("Shape",
AddAttr<std::vector<int>>("shape", "(Tensor<int32>, optional). If provided, reshape according to "
"(vector<int>) " "this given shape. That is to say it has a higher priority than "
"Target shape of reshape operator."); "the shape attribute, while the shape attribute still should be "
"set correctly to gurantee shape inference in compile time.")
.AsDispensable();
AddOutput("Out", "(Tensor). The output tensor of reshape operator.");
AddAttr<std::vector<int>>(
"shape", "(std::vector<int>) Target shape of reshape operator.");
AddAttr<bool>("inplace", AddAttr<bool>("inplace",
"Change the source tensor's shape without copy memory.") "(default: false) Change the source tensor's shape without "
.SetDefault(true); "memory copy. When Attr(inplace) is set true, the output "
"tensor shares memory with Input(X), otherwise, a new output "
"tensor is created, and its data are copied from Input(x).")
.SetDefault(false);
AddComment(R"DOC( AddComment(R"DOC(
Reshape Operator. Reshape Operator.
Reshape Input(X) into the shape specified by Attr(shape). Reshape Input(X) into the shape specified by Attr(shape) or Input(Shape). The
data in Input(X) are unchanged.
Examples:
An example: 1. Given a 3-D tensor Input(X) with a shape [2, 4, 6], and the target shape
Given a 2-D tensor X with 2 rows and 2 columns : [[1, 2], [3, 4]] specified by Attr(shape) is [6, 8], the reshape operator will transform Input(X)
into a 2-D tensor with shape [6, 8] and leaving Input(X)'s data unchanged.
and target shape = [1, 4], the reshape operator will transform 2. Given a 3-D tensor Input(X) with a shape [2, 4, 6], and the target shape
the tensor X into a 2-D tensor: [[1, 2, 3, 4]] specified by Attr(shape) is [2, 3, -1, 2], the reshape operator will transform
Input(X) into a 4-D tensor with shape [2, 3, 4, 2] and leaving Input(X)'s data
unchanged. In this case, one and only dimension of Attr(shape) can be set to -1,
the value of this dimension is inferred from the total element number of
Input(X) and remaining dimensions.
3. Given a 3-D tensor Input(X) with a shape [2, 4, 6], and the target shape
specified by Attr(shape) is [-1, 0, 3, 2], the reshape operator will transform
Input(X) into a 4-D tensor with shape [2, 4, 3, 2] and leaving Input(X)'s data
unchanged. In this case, besides -1, 0 means the actual dimension value is going
to be copied from the corresponding dimension of Input(X).
Note:
1. One and only one dimension in Attr(shape) can be set -1. In this case,
the actual dimension value will be infered from the total element number of
Input(X) and remaining dimensions.
2. More than one dimensions in Attr(shape) can be set to 0, which means the real
dimension value will be copied from Input(X) at runtime. Note that the index of
0 can not exceed Rank(X). For example, Input(X) is a 3-D tensor with shape
[2, 3, 4], Attr(shape) = [2, 3, 2, 0] is an invalid input.
3. Input(Shape) has a higher priority than Attr(shape) if it is provided, while
Attr(shape) still should be set correctly to gurantee shape inference in
compile-time.
One dimension in the target shape can be set -1, representing that its
size is unknown. In this case, the real dimension will be infered from
the original shape of Input(X) and other dimensions in the target shape.
)DOC"); )DOC");
} }
}; };
...@@ -119,6 +95,14 @@ class ReshapeGradOp : public framework::OperatorWithKernel { ...@@ -119,6 +95,14 @@ class ReshapeGradOp : public framework::OperatorWithKernel {
"Input(Out@GRAD) shouldn't be null."); "Input(Out@GRAD) shouldn't be null.");
ctx->SetOutputDim(framework::GradVarName("X"), ctx->GetInputDim("X")); ctx->SetOutputDim(framework::GradVarName("X"), ctx->GetInputDim("X"));
} }
protected:
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext &ctx) const override {
return framework::OpKernelType(
framework::ToDataType(ctx.Input<framework::LoDTensor>("X")->type()),
ctx.device_context());
}
}; };
} // namespace operators } // namespace operators
......
paddle/fluid/operators/reshape_op.h
浏览文件 @ 1f6e0448
...@@ -20,17 +20,129 @@ limitations under the License. */ ...@@ -20,17 +20,129 @@ limitations under the License. */
namespace paddle { namespace paddle {
namespace operators { namespace operators {
class ReshapeOp : public framework::OperatorWithKernel {
public:
ReshapeOp(const std::string &type, const framework::VariableNameMap &inputs,
const framework::VariableNameMap &outputs,
const framework::AttributeMap &attrs)
: OperatorWithKernel(type, inputs, outputs, attrs) {}
void InferShape(framework::InferShapeContext *ctx) const override {
PADDLE_ENFORCE(ctx->HasInput("X"),
"Input(X) of ReshapeOp should not be null.");
PADDLE_ENFORCE(ctx->HasOutput("Out"),
"Output(Out) of ReshapeOp should not be null.");
const std::vector<int> &shape = ctx->Attrs().Get<std::vector<int>>("shape");
PADDLE_ENFORCE(!shape.empty(),
"The shape information must be set by Attr(shape).");
if (ctx->HasInput("Shape") && ctx->IsRuntime()) {
// If true, set the shape of Output(Out) according to Input(Shape) in
// ReshapeKernel with ExecutionContext. Also check LoD in ReshapeKernel.
ctx->ShareLoD("X", /*->*/ "Out");
return;
}
auto x_dims = ctx->GetInputDim("X");
auto out_dims = ValidateShape(shape, x_dims);
ctx->SetOutputDim("Out", out_dims);
if (x_dims[0] == out_dims[0]) {
// Only pass LoD when the first dimension of output and Input(X)
// are the same.
ctx->ShareLoD("X", /*->*/ "Out");
}
}
static framework::DDim ValidateShape(const std::vector<int> shape,
const framework::DDim &in_dims) {
const int64_t in_size = framework::product(in_dims);
// only one dimension canbe set to -1, whose size will be automatically
// infered.
const int64_t unk_dim_val = -1;
const int64_t copy_dim_val = 0;
std::vector<int64_t> output_shape(shape.size(), 0);
int64_t capacity = 1;
int unk_dim_idx = -1;
for (size_t i = 0; i < shape.size(); ++i) {
if (shape[i] == unk_dim_val) {
PADDLE_ENFORCE(
unk_dim_idx == -1,
"Only one input dimension of Attr(shape) can be unknown.");
unk_dim_idx = i;
} else if (shape[i] == copy_dim_val) {
PADDLE_ENFORCE(
static_cast<int>(i) < in_dims.size(),
"The index of dimension to copy from input shape must be less "
"than the size of input shape.");
} else {
PADDLE_ENFORCE(
shape[i] > 0,
"Each input dimension of Attr(shape) must not be negtive except "
"one unknown dimension.");
}
capacity *= (shape[i] ? shape[i] : in_dims[i]);
output_shape[i] =
(shape[i] ? static_cast<int64_t>(shape[i]) : in_dims[i]);
}
if (unk_dim_idx != -1) {
output_shape[unk_dim_idx] = -in_size / capacity;
PADDLE_ENFORCE_EQ(output_shape[unk_dim_idx] * capacity, -in_size,
"Invalid shape is given.");
} else {
PADDLE_ENFORCE_EQ(capacity, in_size, "Invalid shape is given.");
}
return framework::make_ddim(output_shape);
}
protected:
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext &ctx) const override {
return framework::OpKernelType(
framework::ToDataType(ctx.Input<framework::LoDTensor>("X")->type()),
ctx.device_context());
}
};
template <typename DeviceContext, typename T> template <typename DeviceContext, typename T>
class ReshapeKernel : public framework::OpKernel<T> { class ReshapeKernel : public framework::OpKernel<T> {
public: public:
void Compute(const framework::ExecutionContext& ctx) const { void Compute(const framework::ExecutionContext &ctx) const {
auto* out = ctx.Output<framework::Tensor>("Out"); auto *out = ctx.Output<framework::LoDTensor>("Out");
auto* in = ctx.Input<framework::Tensor>("X"); auto *in = ctx.Input<framework::LoDTensor>("X");
auto *shape_tensor = ctx.Input<framework::LoDTensor>("Shape");
framework::DDim out_dims = out->dims();
if (shape_tensor) {
auto *shape_data = shape_tensor->data<int>();
if (platform::is_gpu_place(ctx.GetPlace())) {
framework::Tensor cpu_shape_tensor;
TensorCopy(*shape_tensor, platform::CPUPlace(), ctx.device_context(),
&cpu_shape_tensor);
shape_data = cpu_shape_tensor.data<int>();
}
auto shape =
std::vector<int>(shape_data, shape_data + shape_tensor->numel());
out_dims = ReshapeOp::ValidateShape(shape, in->dims());
}
if (!in->lod().empty()) {
PADDLE_ENFORCE_EQ(
out_dims[0], in->dims()[0],
"Reshape operator cannot reshape an input sequence batch "
"into an output sequence batch that has a different "
"number of time steps. Please consider using "
"sequence_reshape op.");
}
bool inplace = ctx.Attr<bool>("inplace"); bool inplace = ctx.Attr<bool>("inplace");
auto out_dims = out->dims(); out->Resize(out_dims);
if (!inplace) { if (!inplace) {
out->mutable_data<T>(ctx.GetPlace()); out->mutable_data<T>(ctx.GetPlace());
framework::TensorCopy(*in, ctx.GetPlace(), ctx.device_context(), out); framework::TensorCopy(*in, ctx.GetPlace(), ctx.device_context(), out);
// TensorCopy will resize to in_dims.
out->Resize(out_dims); out->Resize(out_dims);
} else { } else {
out->ShareDataWith(*in); out->ShareDataWith(*in);
...@@ -42,9 +154,10 @@ class ReshapeKernel : public framework::OpKernel<T> { ...@@ -42,9 +154,10 @@ class ReshapeKernel : public framework::OpKernel<T> {
template <typename DeviceContext, typename T> template <typename DeviceContext, typename T>
class ReshapeGradKernel : public framework::OpKernel<T> { class ReshapeGradKernel : public framework::OpKernel<T> {
public: public:
void Compute(const framework::ExecutionContext& ctx) const { void Compute(const framework::ExecutionContext &ctx) const {
auto* d_out = ctx.Input<framework::Tensor>(framework::GradVarName("Out")); auto *d_out = ctx.Input<framework::Tensor>(framework::GradVarName("Out"));
auto* d_x = ctx.Output<framework::Tensor>(framework::GradVarName("X")); auto *d_x = ctx.Output<framework::Tensor>(framework::GradVarName("X"));
d_x->mutable_data<T>(ctx.GetPlace()); d_x->mutable_data<T>(ctx.GetPlace());
bool inplace = ctx.Attr<bool>("inplace"); bool inplace = ctx.Attr<bool>("inplace");
......
paddle/fluid/pybind/pybind.cc
浏览文件 @ 1f6e0448
...@@ -504,10 +504,10 @@ All parameter, weight, gradient are variables in Paddle. ...@@ -504,10 +504,10 @@ All parameter, weight, gradient are variables in Paddle.
const std::unordered_set<std::string> &params, const std::unordered_set<std::string> &params,
const ProgramDesc &startup_program, const ProgramDesc &startup_program,
const ProgramDesc &main_program, const std::string &loss_var_name, const ProgramDesc &main_program, const std::string &loss_var_name,
Scope *scope) { Scope *scope, bool allow_op_delay) {
new (&self) ParallelExecutor(num_threads, use_event, places, new (&self) ParallelExecutor(num_threads, use_event, places,
params, startup_program, main_program, params, startup_program, main_program,
loss_var_name, scope); loss_var_name, scope, allow_op_delay);
}) })
.def("run", &ParallelExecutor::Run); .def("run", &ParallelExecutor::Run);
......
paddle/scripts/docker/build.sh
浏览文件 @ 1f6e0448
...@@ -104,7 +104,9 @@ EOF ...@@ -104,7 +104,9 @@ EOF
# make install should also be test when unittest # make install should also be test when unittest
make install -j `nproc` make install -j `nproc`
pip install /usr/local/opt/paddle/share/wheels/*.whl pip install /usr/local/opt/paddle/share/wheels/*.whl
paddle version if [[ ${WITH_FLUID_ONLY:-OFF} == "OFF" ]] ; then
paddle version
fi
fi fi
} }
...@@ -183,6 +185,14 @@ EOF ...@@ -183,6 +185,14 @@ EOF
NCCL_DEPS="" NCCL_DEPS=""
fi fi
if [[ ${WITH_FLUID_ONLY:-OFF} == "OFF" ]]; then
PADDLE_VERSION="paddle version"
CMD='"paddle", "version"'
else
PADDLE_VERSION="true"
CMD='"true"'
fi
cat >> /paddle/build/Dockerfile <<EOF cat >> /paddle/build/Dockerfile <<EOF
ADD python/dist/*.whl / ADD python/dist/*.whl /
# run paddle version to install python packages first # run paddle version to install python packages first
...@@ -192,7 +202,7 @@ EOF ...@@ -192,7 +202,7 @@ EOF
pip install /*.whl; apt-get install -f -y && \ pip install /*.whl; apt-get install -f -y && \
apt-get clean -y && \ apt-get clean -y && \
rm -f /*.whl && \ rm -f /*.whl && \
paddle version && \ ${PADDLE_VERSION} && \
ldconfig ldconfig
${DOCKERFILE_CUDNN_DSO} ${DOCKERFILE_CUDNN_DSO}
${DOCKERFILE_GPU_ENV} ${DOCKERFILE_GPU_ENV}
...@@ -200,7 +210,7 @@ EOF ...@@ -200,7 +210,7 @@ EOF
ADD go/cmd/pserver/pserver /usr/bin/ ADD go/cmd/pserver/pserver /usr/bin/
ADD go/cmd/master/master /usr/bin/ ADD go/cmd/master/master /usr/bin/
# default command shows the paddle version and exit # default command shows the paddle version and exit
CMD ["paddle", "version"] CMD [${CMD}]
EOF EOF
} }
......
python/CMakeLists.txt
浏览文件 @ 1f6e0448
...@@ -81,6 +81,7 @@ if (WITH_TESTING) ...@@ -81,6 +81,7 @@ if (WITH_TESTING)
# enable v2 API unittest only when paddle swig api is compiled # enable v2 API unittest only when paddle swig api is compiled
add_subdirectory(paddle/v2/tests) add_subdirectory(paddle/v2/tests)
add_subdirectory(paddle/v2/plot/tests) add_subdirectory(paddle/v2/plot/tests)
add_subdirectory(paddle/v2/reader/tests)
endif() endif()
endif() endif()
add_subdirectory(paddle/fluid/tests) add_subdirectory(paddle/fluid/tests)
......
python/paddle/dataset/__init__.py
浏览文件 @ 1f6e0448
...@@ -37,7 +37,7 @@ __all__ = [ ...@@ -37,7 +37,7 @@ __all__ = [
'cifar', 'cifar',
'movielens', 'movielens',
'conll05', 'conll05',
'sentiment' 'sentiment',
'uci_housing', 'uci_housing',
'wmt14', 'wmt14',
'wmt16', 'wmt16',
......
python/paddle/fluid/distribute_transpiler.py
浏览文件 @ 1f6e0448
...@@ -276,20 +276,25 @@ class DistributeTranspiler: ...@@ -276,20 +276,25 @@ class DistributeTranspiler:
suff_idx = v.name.find(".trainer_") suff_idx = v.name.find(".trainer_")
if suff_idx >= 0: if suff_idx >= 0:
orig_var_name = v.name[:suff_idx] orig_var_name = v.name[:suff_idx]
pserver_program.global_block().create_var( else:
orig_var_name = v.name
single_trainer_var = pserver_program.global_block().create_var(
name=orig_var_name, name=orig_var_name,
persistable=True, persistable=True,
type=v.type, type=v.type,
dtype=v.dtype, dtype=v.dtype,
shape=v.shape) shape=v.shape)
for trainer_id in xrange(self.trainers): if self.trainers > 1:
var = pserver_program.global_block().create_var( for trainer_id in xrange(self.trainers):
name="%s.trainer_%d" % (orig_var_name, trainer_id), var = pserver_program.global_block().create_var(
persistable=False, name="%s.trainer_%d" % (orig_var_name, trainer_id),
type=v.type, persistable=False,
dtype=v.dtype, type=v.type,
shape=v.shape) dtype=v.dtype,
recv_inputs.append(var) shape=v.shape)
recv_inputs.append(var)
else:
recv_inputs.append(single_trainer_var)
# step3 # step3
optimize_block = pserver_program.create_block(0) optimize_block = pserver_program.create_block(0)
...@@ -511,8 +516,11 @@ class DistributeTranspiler: ...@@ -511,8 +516,11 @@ class DistributeTranspiler:
def _append_split_op(self, program, gradblocks): def _append_split_op(self, program, gradblocks):
# Split variables that need to be split and append respective ops # Split variables that need to be split and append respective ops
add_suffix = False
if self.trainers > 1:
add_suffix = True
var_mapping = self._create_vars_from_blocklist( var_mapping = self._create_vars_from_blocklist(
program, gradblocks, add_trainer_suffix=True) program, gradblocks, add_trainer_suffix=add_suffix)
for varname, splited_vars in var_mapping.iteritems(): for varname, splited_vars in var_mapping.iteritems():
# variable that don't need to split have empty splited_vars # variable that don't need to split have empty splited_vars
if len(splited_vars) <= 1: if len(splited_vars) <= 1:
......
python/paddle/fluid/executor.py
浏览文件 @ 1f6e0448
...@@ -48,8 +48,7 @@ def as_numpy(tensor): ...@@ -48,8 +48,7 @@ def as_numpy(tensor):
assert isinstance(tensor, core.LoDTensor) assert isinstance(tensor, core.LoDTensor)
lod = tensor.lod() lod = tensor.lod()
if len(lod) > 0: if len(lod) > 0:
raise RuntimeError( raise RuntimeError("Some of your fetched tensors hold LoD information. \
"Some of your featched tensors hold LoD information. \
They can not be completely cast to Python ndarray. \ They can not be completely cast to Python ndarray. \
Please set the parameter 'return_numpy' as 'False' to \ Please set the parameter 'return_numpy' as 'False' to \
return LoDTensor itself directly.") return LoDTensor itself directly.")
...@@ -180,60 +179,24 @@ def get_program_cache_key(feed, fetch_list): ...@@ -180,60 +179,24 @@ def get_program_cache_key(feed, fetch_list):
class Executor(object): class Executor(object):
def __init__(self, places): def __init__(self, place):
if not isinstance(places, list) and not isinstance(places, tuple): self.place = place
places = [places] p = core.Place()
p.set_place(place)
act_places = [] self.executor = core.Executor(p)
for each in places:
p = core.Place()
p.set_place(each)
act_places.append(p)
# TODO(dzhwinter) : only use the first place
self.executor = core.Executor(act_places[0])
self.places = places
self.program_caches = dict() self.program_caches = dict()
def aslodtensor(self, data): def as_lodtensor(self, data):
def accumulate(data): if isinstance(data, list):
if not isinstance(data, list): raise RuntimeError("Some of your feed data hold LoD information. \
return 1 They can not be completely cast from a list of Python \
return sum([accumulate(sub) for sub in data]) ndarray to LoDTensor. Please convert data to LoDTensor \
directly before feeding the data.\
def parselod(data): ")
seq_lens = [accumulate(seq) for seq in data] # single tensor case
cur_len = 0 tensor = core.LoDTensor()
lod = [cur_len] tensor.set(data, self.place)
for l in seq_lens: return tensor
cur_len += l
lod.append(cur_len)
return lod
assert len(self.places) != 0
if not isinstance(data, list):
# pure tensor case
tensor = core.LoDTensor()
tensor.set(data, self.places[0])
return tensor
else:
raise RuntimeError("Current implementation lacks unittests")
# lodtensor case
lod = []
if not isinstance(data[0], list):
lod.append(parselod(data))
flattened_data = np.concatenate(data, axis=0).astype("int64")
else:
while isinstance(data[0], list):
lod.append(parselod(seq))
flattened_data = [item for seq in data for item in seq]
data = flattened_data
flattened_data = np.concatenate(data, axis=0).astype("int64")
flattened_data = flattened_data.reshape([len(flattened_data), 1])
tensor = core.LoDTensor()
tensor.set(flattened_data, self.places[0])
tensor.set_lod(lod)
return tensor
def _get_program_cache(self, program_cache_key): def _get_program_cache(self, program_cache_key):
return self.program_caches.get(program_cache_key, None) return self.program_caches.get(program_cache_key, None)
...@@ -293,7 +256,7 @@ class Executor(object): ...@@ -293,7 +256,7 @@ class Executor(object):
feed_target_name = op.desc.output('Out')[0] feed_target_name = op.desc.output('Out')[0]
cur_feed = feed[feed_target_name] cur_feed = feed[feed_target_name]
if not isinstance(cur_feed, core.LoDTensor): if not isinstance(cur_feed, core.LoDTensor):
cur_feed = self.aslodtensor(cur_feed) cur_feed = self.as_lodtensor(cur_feed)
idx = op.desc.attr('col') idx = op.desc.attr('col')
core.set_feed_variable(scope, cur_feed, feed_var_name, idx) core.set_feed_variable(scope, cur_feed, feed_var_name, idx)
else: else:
......
python/paddle/fluid/layers/detection.py
浏览文件 @ 1f6e0448
...@@ -19,7 +19,6 @@ from layer_function_generator import generate_layer_fn ...@@ -19,7 +19,6 @@ from layer_function_generator import generate_layer_fn
from layer_function_generator import autodoc from layer_function_generator import autodoc
from ..layer_helper import LayerHelper from ..layer_helper import LayerHelper
import tensor import tensor
import ops
import nn import nn
import math import math
...@@ -58,7 +57,7 @@ def detection_output(loc, ...@@ -58,7 +57,7 @@ def detection_output(loc,
This operation is to get the detection results by performing following This operation is to get the detection results by performing following
two steps: two steps:
1. Decode input bounding box predictions according to the prior boxes. 1. Decode input bounding box predictions according to the prior boxes.
2. Get the final detection results by applying multi-class non maximum 2. Get the final detection results by applying multi-class non maximum
suppression (NMS). suppression (NMS).
...@@ -130,9 +129,9 @@ def detection_output(loc, ...@@ -130,9 +129,9 @@ def detection_output(loc,
target_box=loc, target_box=loc,
code_type='decode_center_size') code_type='decode_center_size')
old_shape = scores.shape old_shape = scores.shape
scores = ops.reshape(x=scores, shape=(-1, old_shape[-1])) scores = nn.reshape(x=scores, shape=(-1, old_shape[-1]))
scores = nn.softmax(input=scores) scores = nn.softmax(input=scores)
scores = ops.reshape(x=scores, shape=old_shape) scores = nn.reshape(x=scores, shape=old_shape)
scores = nn.transpose(scores, perm=[0, 2, 1]) scores = nn.transpose(scores, perm=[0, 2, 1])
scores.stop_gradient = True scores.stop_gradient = True
nmsed_outs = helper.create_tmp_variable(dtype=decoded_box.dtype) nmsed_outs = helper.create_tmp_variable(dtype=decoded_box.dtype)
...@@ -463,7 +462,7 @@ def ssd_loss(location, ...@@ -463,7 +462,7 @@ def ssd_loss(location,
num, num_prior, num_class = confidence.shape num, num_prior, num_class = confidence.shape
def __reshape_to_2d(var): def __reshape_to_2d(var):
return ops.reshape(x=var, shape=[-1, var.shape[-1]]) return nn.reshape(x=var, shape=[-1, var.shape[-1]])
# 1. Find matched boundding box by prior box. # 1. Find matched boundding box by prior box.
# 1.1 Compute IOU similarity between ground-truth boxes and prior boxes. # 1.1 Compute IOU similarity between ground-truth boxes and prior boxes.
...@@ -474,7 +473,7 @@ def ssd_loss(location, ...@@ -474,7 +473,7 @@ def ssd_loss(location,
# 2. Compute confidence for mining hard examples # 2. Compute confidence for mining hard examples
# 2.1. Get the target label based on matched indices # 2.1. Get the target label based on matched indices
gt_label = ops.reshape(x=gt_label, shape=gt_label.shape + (1, )) gt_label = nn.reshape(x=gt_label, shape=gt_label.shape + (1, ))
gt_label.stop_gradient = True gt_label.stop_gradient = True
target_label, _ = target_assign( target_label, _ = target_assign(
gt_label, matched_indices, mismatch_value=background_label) gt_label, matched_indices, mismatch_value=background_label)
...@@ -487,7 +486,7 @@ def ssd_loss(location, ...@@ -487,7 +486,7 @@ def ssd_loss(location,
conf_loss = nn.softmax_with_cross_entropy(confidence, target_label) conf_loss = nn.softmax_with_cross_entropy(confidence, target_label)
# 3. Mining hard examples # 3. Mining hard examples
conf_loss = ops.reshape(x=conf_loss, shape=(num, num_prior)) conf_loss = nn.reshape(x=conf_loss, shape=(num, num_prior))
conf_loss.stop_gradient = True conf_loss.stop_gradient = True
neg_indices = helper.create_tmp_variable(dtype='int32') neg_indices = helper.create_tmp_variable(dtype='int32')
dtype = matched_indices.dtype dtype = matched_indices.dtype
...@@ -556,7 +555,7 @@ def ssd_loss(location, ...@@ -556,7 +555,7 @@ def ssd_loss(location,
# 5.3 Compute overall weighted loss. # 5.3 Compute overall weighted loss.
loss = conf_loss_weight * conf_loss + loc_loss_weight * loc_loss loss = conf_loss_weight * conf_loss + loc_loss_weight * loc_loss
# reshape to [N, Np], N is the batch size and Np is the prior box number. # reshape to [N, Np], N is the batch size and Np is the prior box number.
loss = ops.reshape(x=loss, shape=[-1, num_prior]) loss = nn.reshape(x=loss, shape=[-1, num_prior])
loss = nn.reduce_sum(loss, dim=1, keep_dim=True) loss = nn.reduce_sum(loss, dim=1, keep_dim=True)
if normalize: if normalize:
normalizer = nn.reduce_sum(target_loc_weight) normalizer = nn.reduce_sum(target_loc_weight)
...@@ -709,7 +708,7 @@ def multi_box_head(inputs, ...@@ -709,7 +708,7 @@ def multi_box_head(inputs,
new_shape = [ new_shape = [
-1, reduce(lambda x, y: x * y, input.shape[axis:len(input.shape)]) -1, reduce(lambda x, y: x * y, input.shape[axis:len(input.shape)])
] ]
out = ops.reshape(x=input, shape=new_shape) out = nn.reshape(x=input, shape=new_shape)
return out return out
def _is_list_or_tuple_(data): def _is_list_or_tuple_(data):
...@@ -803,7 +802,7 @@ def multi_box_head(inputs, ...@@ -803,7 +802,7 @@ def multi_box_head(inputs,
mbox_loc.shape[0], mbox_loc.shape[0],
mbox_loc.shape[1] * mbox_loc.shape[2] * mbox_loc.shape[3] / 4, 4 mbox_loc.shape[1] * mbox_loc.shape[2] * mbox_loc.shape[3] / 4, 4
] ]
mbox_loc_flatten = ops.reshape(mbox_loc, shape=new_shape) mbox_loc_flatten = nn.reshape(mbox_loc, shape=new_shape)
mbox_locs.append(mbox_loc_flatten) mbox_locs.append(mbox_loc_flatten)
# get conf # get conf
...@@ -819,7 +818,7 @@ def multi_box_head(inputs, ...@@ -819,7 +818,7 @@ def multi_box_head(inputs,
conf_loc.shape[0], conf_loc.shape[1] * conf_loc.shape[2] * conf_loc.shape[0], conf_loc.shape[1] * conf_loc.shape[2] *
conf_loc.shape[3] / num_classes, num_classes conf_loc.shape[3] / num_classes, num_classes
] ]
conf_loc_flatten = ops.reshape(conf_loc, shape=new_shape) conf_loc_flatten = nn.reshape(conf_loc, shape=new_shape)
mbox_confs.append(conf_loc_flatten) mbox_confs.append(conf_loc_flatten)
if len(box_results) == 1: if len(box_results) == 1:
......
python/paddle/fluid/layers/nn.py
浏览文件 @ 1f6e0448
...@@ -73,8 +73,10 @@ __all__ = [ ...@@ -73,8 +73,10 @@ __all__ = [
'smooth_l1', 'smooth_l1',
'one_hot', 'one_hot',
'autoincreased_step_counter', 'autoincreased_step_counter',
'reshape',
'lod_reset', 'lod_reset',
'lrn', 'lrn',
'pad',
] ]
...@@ -3265,6 +3267,8 @@ def one_hot(input, depth): ...@@ -3265,6 +3267,8 @@ def one_hot(input, depth):
The one-hot tensor or LodTensor, same as input. The one-hot tensor or LodTensor, same as input.
Examples: Examples:
.. code-block:: python
X is a LoDTensor: X is a LoDTensor:
X.lod = [[0, 1, 4]] X.lod = [[0, 1, 4]]
X.shape = [4, 1] X.shape = [4, 1]
...@@ -3319,6 +3323,102 @@ def autoincreased_step_counter(counter_name=None, begin=1, step=1): ...@@ -3319,6 +3323,102 @@ def autoincreased_step_counter(counter_name=None, begin=1, step=1):
return counter return counter
def reshape(x, shape, actual_shape=None, act=None, inplace=True, name=None):
"""
Gives a new shape to the input Tensor without changing its data.
The target shape can be given by :attr:`shape` or :attr:`actual_shape`.
:attr:`shape` is a list of integer while :attr:`actual_shape` is a tensor
variable. :attr:`actual_shape` has a higher priority than :attr:`shape`
if it is provided, while :attr:`shape` still should be set correctly to
gurantee shape inference in compile-time.
Some tricks exist when specifying the target shape.
1. -1 means the value of this dimension is inferred from the total element
number of x and remaining dimensions. Thus one and only one dimension can
be set -1.
2. 0 means the actual dimension value is going to be copied from the
corresponding dimension of x. The indice of 0s in shape can not exceed
Rank(X).
Here are some examples to explain it.
1. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
is [6, 8], the reshape operator will transform x into a 2-D tensor with
shape [6, 8] and leaving x's data unchanged.
2. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
specified is [2, 3, -1, 2], the reshape operator will transform x into a
4-D tensor with shape [2, 3, 4, 2] and leaving x's data unchanged. In this
case, one dimension of the target shape is set to -1, the value of this
dimension is inferred from the total element number of x and remaining
dimensions.
3. Given a 3-D tensor x with a shape [2, 4, 6], and the target shape
is [-1, 0, 3, 2], the reshape operator will transform x into a 4-D tensor
with shape [2, 4, 3, 2] and leaving x's data unchanged. In this case,
besides -1, 0 means the actual dimension value is going to be copied from
the corresponding dimension of x.
Args:
input(variable): The input tensor.
shape(list): The new shape. At most one dimension of the new shape can
be -1.
actual_shape(variable): An optional input. If provided, reshape
according to this given shape rather than
:attr:`shape` specifying shape. That is to
say :attr:`actual_shape` has a higher priority
than :attr:`shape`.
act (str): The non-linear activation to be applied to output variable.
inplace(bool): If this flag is set true, a new output tensor is created
whose data is copied from input x, otherwise the output
shares data with input without copying.
Returns(variable): The output tensor.
Examples:
.. code-block:: python
data = fluid.layers.data(
name='data', shape=[2, 4, 6], dtype='float32')
reshaped = fluid.layers.reshape(
x=data, shape=[-1, 0, 3, 2], act='tanh', inplace=True)
"""
if not (isinstance(shape, list) or isinstance(shape, tuple)):
raise ValueError("Input shape must be a python lsit or tuple.")
# Validate the shape
unk_dim_idx = -1
for dim_idx, dim_size in enumerate(shape):
if dim_size == -1:
assert unk_dim_idx == -1, (
"Only one dimension in shape can be unknown.")
unk_dim_idx = dim_idx
elif dim_size == 0:
assert dim_idx < len(x.shape), (
"The indice of 0s in shape can not exceed Rank(X).")
else:
assert dim_size > 0, (
"Each dimension size given in shape must not be negtive "
"except one unknown dimension.")
helper = LayerHelper("reshape", **locals())
reshaped = helper.create_tmp_variable(dtype=x.dtype)
helper.append_op(
type="reshape",
inputs={"X": x,
"Shape": actual_shape}
if isinstance(actual_shape, Variable) else {"X": x},
attrs={"shape": shape,
"inplace": inplace},
outputs={"Out": reshaped})
return helper.append_activation(reshaped)
def lod_reset(x, y=None, target_lod=None): def lod_reset(x, y=None, target_lod=None):
""" """
LoD Reset Operator. Set LoD of **x** to a new one specified by **y** or LoD Reset Operator. Set LoD of **x** to a new one specified by **y** or
...@@ -3482,3 +3582,62 @@ def lrn(input, n=5, k=1.0, alpha=1e-4, beta=0.75, name=None): ...@@ -3482,3 +3582,62 @@ def lrn(input, n=5, k=1.0, alpha=1e-4, beta=0.75, name=None):
"beta": beta}) "beta": beta})
return lrn_out return lrn_out
def pad(x, paddings, pad_value=0., name=None):
"""
Pads a tensor with a constant value given by :attr:`pad_value`, and the
padded width is specified by :attr:`paddings`.
Specifically, the number of values padded before the contents of :attr:`x`
in dimension :attr:`i` is indicated by :attr:`paddings[i]`, and the number
of values padded after the contents of :attr:`x` in dimension :attr:`i` is
indicated by :attr:`paddings[i+1]`.
See below for an example.
.. code-block:: text
Given:
x = [[1, 2], [3, 4]]
paddings = [0, 1, 1, 2]
pad_value = 0
Return:
out = [[0, 1, 2, 0, 0]
[0, 3, 4, 0, 0]
[0, 0, 0, 0, 0]]
Args:
x (Variable): The input tensor variable.
paddings (list): A list of integers. Its elements specify the padded
width before and after for each dimension in turn.
The length of :attr:paddings must be
:math:`rank(x) \\times 2`.
pad_value (float): The constant value used to pad.
name(str|None): A name for this layer(optional). If set None, the layer
will be named automatically.
Returns:
Variable: The padded tensor variable.
Examples:
.. code-block:: python
# x is a rank 2 tensor variable.
out = fluid.layers.pad(
x=x, paddings=[0, 1, 1, 2], pad_value=0.)
"""
helper = LayerHelper('pad', input=x, **locals())
dtype = helper.input_dtype()
out = helper.create_tmp_variable(dtype)
helper.append_op(
type='pad',
inputs={'X': x},
outputs={'Out': out},
attrs={'paddings': paddings,
'pad_value': float(pad_value)})
return out
python/paddle/fluid/layers/ops.py
浏览文件 @ 1f6e0448
...@@ -49,7 +49,6 @@ __activations__ = [ ...@@ -49,7 +49,6 @@ __activations__ = [
__all__ = [ __all__ = [
'mean', 'mean',
'mul', 'mul',
'reshape',
'scale', 'scale',
'sigmoid_cross_entropy_with_logits', 'sigmoid_cross_entropy_with_logits',
'elementwise_add', 'elementwise_add',
......
python/paddle/fluid/parallel_executor.py
浏览文件 @ 1f6e0448
...@@ -21,7 +21,11 @@ __all__ = ['ParallelExecutor'] ...@@ -21,7 +21,11 @@ __all__ = ['ParallelExecutor']
class ParallelExecutor(object): class ParallelExecutor(object):
def __init__(self, loss_name, use_cuda, num_threads=None): def __init__(self,
loss_name,
use_cuda,
num_threads=None,
allow_op_delay=False):
places = [] places = []
if use_cuda: if use_cuda:
for i in xrange(core.get_cuda_device_count()): for i in xrange(core.get_cuda_device_count()):
...@@ -35,7 +39,12 @@ class ParallelExecutor(object): ...@@ -35,7 +39,12 @@ class ParallelExecutor(object):
places.append(p) places.append(p)
if num_threads is None: if num_threads is None:
num_threads = min(len(places) * 2, multiprocessing.cpu_count()) if use_cuda:
# Experiments on se-resnext shows that too many threads hurt
# performance. Worth tunning for other models in the future.
num_threads = len(places)
else:
min(len(places) * 2, multiprocessing.cpu_count())
startup = framework.default_startup_program() startup = framework.default_startup_program()
main = framework.default_main_program() main = framework.default_main_program()
...@@ -52,7 +61,8 @@ class ParallelExecutor(object): ...@@ -52,7 +61,8 @@ class ParallelExecutor(object):
startup.desc, startup.desc,
main.desc, main.desc,
loss_name, loss_name,
scope) scope,
allow_op_delay)
self.scope = scope self.scope = scope
def run(self, fetch_list): def run(self, fetch_list):
......
python/paddle/fluid/tests/unittests/CMakeLists.txt
浏览文件 @ 1f6e0448
...@@ -29,6 +29,7 @@ function(py_test_modules TARGET_NAME) ...@@ -29,6 +29,7 @@ function(py_test_modules TARGET_NAME)
endfunction() endfunction()
# test time consuming OPs in a separate process for expliot parallism # test time consuming OPs in a separate process for expliot parallism
list(REMOVE_ITEM TEST_OPS test_parallel_executor)
list(REMOVE_ITEM TEST_OPS test_warpctc_op) list(REMOVE_ITEM TEST_OPS test_warpctc_op)
list(REMOVE_ITEM TEST_OPS test_dyn_rnn) list(REMOVE_ITEM TEST_OPS test_dyn_rnn)
list(REMOVE_ITEM TEST_OPS test_mul_op) list(REMOVE_ITEM TEST_OPS test_mul_op)
...@@ -64,6 +65,7 @@ else() ...@@ -64,6 +65,7 @@ else()
endif(WITH_FAST_BUNDLE_TEST) endif(WITH_FAST_BUNDLE_TEST)
# tests with high overhead # tests with high overhead
py_test_modules(test_parallel_executor MODULES test_parallel_executor)
py_test_modules(test_warpctc_op MODULES test_warpctc_op ENVS FLAGS_warpctc_dir=${WARPCTC_LIB_DIR}) py_test_modules(test_warpctc_op MODULES test_warpctc_op ENVS FLAGS_warpctc_dir=${WARPCTC_LIB_DIR})
py_test_modules(test_train_dyn_rnn MODULES test_dyn_rnn) py_test_modules(test_train_dyn_rnn MODULES test_dyn_rnn)
py_test_modules(test_mul_op MODULES test_mul_op) py_test_modules(test_mul_op MODULES test_mul_op)
......
python/paddle/fluid/tests/unittests/op_test.py
浏览文件 @ 1f6e0448
...@@ -334,7 +334,7 @@ class OpTest(unittest.TestCase): ...@@ -334,7 +334,7 @@ class OpTest(unittest.TestCase):
np.allclose( np.allclose(
actual_t, expect_t, atol=atol), actual_t, expect_t, atol=atol),
"Output (" + out_name + ") has diff at " + str(place) + "Output (" + out_name + ") has diff at " + str(place) +
str(actual_t) + str(expect_t)) str(actual_t) + "\n" + str(expect_t))
if isinstance(expect, tuple): if isinstance(expect, tuple):
self.assertListEqual(actual.lod(), expect[1], self.assertListEqual(actual.lod(), expect[1],
"Output (" + out_name + "Output (" + out_name +
...@@ -568,6 +568,6 @@ class OpTest(unittest.TestCase): ...@@ -568,6 +568,6 @@ class OpTest(unittest.TestCase):
fetch_list = [g for p, g in param_grad_list] fetch_list = [g for p, g in param_grad_list]
executor = Executor(place) executor = Executor(place)
return map( return map(np.array,
np.array, executor.run(prog, feed_dict, fetch_list,
executor.run(prog, feed_dict, fetch_list, return_numpy=False)) return_numpy=False))
python/paddle/fluid/tests/unittests/test_lookup_table_op.py
浏览文件 @ 1f6e0448
...@@ -96,5 +96,47 @@ class TestLookupTableIdsIsSelectedRows(OpTest): ...@@ -96,5 +96,47 @@ class TestLookupTableIdsIsSelectedRows(OpTest):
self.check_with_place(place) self.check_with_place(place)
class TestLookupTableWIsSelectedRows(OpTest):
def check_with_place(self, place):
scope = core.Scope()
# create and initialize Id Variable
ids_tensor = scope.var('Ids').get_tensor()
ids_array = np.array([[0], [4], [3], [5]]).astype("int64")
ids_tensor.set(ids_array, place)
# create and initialize W Variable
rows = [0, 1, 2, 3, 4, 5, 6]
row_numel = 12
w_selected_rows = scope.var('W').get_selected_rows()
w_selected_rows.set_height(len(rows))
w_selected_rows.set_rows(rows)
w_array = np.ones((len(rows), row_numel)).astype("float32")
for i in range(len(rows)):
w_array[i] *= i
ids_tensor = w_selected_rows.get_tensor()
ids_tensor.set(w_array, place)
# create Out Variable
Out_tensor = scope.var('Out').get_tensor()
# create and run lookup_table operator
lookup_table = Operator("lookup_table", W='W', Ids='Ids', Out='Out')
lookup_table.run(scope, place)
# get result from Out
result_array = np.array(Out_tensor)
# all(): return True if all elements of the iterable are true (or if the iterable is empty)
for idx, row in enumerate(ids_array):
assert (row[0] == result_array[idx]).all()
def test_w_is_selected_rows(self):
places = [core.CPUPlace()]
# currently only support CPU
for place in places:
self.check_with_place(place)
if __name__ == "__main__": if __name__ == "__main__":
unittest.main() unittest.main()
python/paddle/fluid/tests/unittests/test_parallel_executor.py
浏览文件 @ 1f6e0448
...@@ -135,18 +135,18 @@ def bottleneck_block(input, num_filters, stride, cardinality, reduction_ratio): ...@@ -135,18 +135,18 @@ def bottleneck_block(input, num_filters, stride, cardinality, reduction_ratio):
return fluid.layers.elementwise_add(x=short, y=scale, act='relu') return fluid.layers.elementwise_add(x=short, y=scale, act='relu')
def SE_ResNeXt152(batch_size=4): def SE_ResNeXt152Small(batch_size=2):
img = fluid.layers.fill_constant( img = fluid.layers.fill_constant(
shape=[batch_size, 3, 224, 224], dtype='float32', value=0.0) shape=[batch_size, 3, 224, 224], dtype='float32', value=0.0)
label = fluid.layers.fill_constant( label = fluid.layers.fill_constant(
shape=[batch_size, 1], dtype='int64', value=0.0) shape=[batch_size, 1], dtype='int64', value=0.0)
conv = conv_bn_layer( conv = conv_bn_layer(
input=img, num_filters=64, filter_size=3, stride=2, act='relu') input=img, num_filters=16, filter_size=3, stride=2, act='relu')
conv = conv_bn_layer( conv = conv_bn_layer(
input=conv, num_filters=64, filter_size=3, stride=1, act='relu') input=conv, num_filters=16, filter_size=3, stride=1, act='relu')
conv = conv_bn_layer( conv = conv_bn_layer(
input=conv, num_filters=128, filter_size=3, stride=1, act='relu') input=conv, num_filters=16, filter_size=3, stride=1, act='relu')
conv = fluid.layers.pool2d( conv = fluid.layers.pool2d(
input=conv, pool_size=3, pool_stride=2, pool_padding=1, pool_type='max') input=conv, pool_size=3, pool_stride=2, pool_padding=1, pool_type='max')
...@@ -184,7 +184,8 @@ class TestParallelExecutorBase(unittest.TestCase): ...@@ -184,7 +184,8 @@ class TestParallelExecutorBase(unittest.TestCase):
method, method,
memory_opt=True, memory_opt=True,
iter=10, iter=10,
batch_size=None): batch_size=None,
allow_op_delay=False):
main = fluid.Program() main = fluid.Program()
startup = fluid.Program() startup = fluid.Program()
with fluid.program_guard(main, startup): with fluid.program_guard(main, startup):
...@@ -194,7 +195,10 @@ class TestParallelExecutorBase(unittest.TestCase): ...@@ -194,7 +195,10 @@ class TestParallelExecutorBase(unittest.TestCase):
if memory_opt: if memory_opt:
fluid.memory_optimize(main) fluid.memory_optimize(main)
exe = fluid.ParallelExecutor(loss_name=loss.name, use_cuda=True) exe = fluid.ParallelExecutor(
loss_name=loss.name,
use_cuda=True,
allow_op_delay=allow_op_delay)
if batch_size is not None: if batch_size is not None:
batch_size *= fluid.core.get_cuda_device_count() batch_size *= fluid.core.get_cuda_device_count()
begin = time.time() begin = time.time()
...@@ -222,7 +226,7 @@ class TestMNIST(TestParallelExecutorBase): ...@@ -222,7 +226,7 @@ class TestMNIST(TestParallelExecutorBase):
def setUpClass(cls): def setUpClass(cls):
# Convert mnist to recordio file # Convert mnist to recordio file
with fluid.program_guard(fluid.Program(), fluid.Program()): with fluid.program_guard(fluid.Program(), fluid.Program()):
reader = paddle.batch(mnist.train(), batch_size=32) reader = paddle.batch(mnist.train(), batch_size=4)
feeder = fluid.DataFeeder( feeder = fluid.DataFeeder(
feed_list=[ # order is image and label feed_list=[ # order is image and label
fluid.layers.data( fluid.layers.data(
...@@ -236,9 +240,11 @@ class TestMNIST(TestParallelExecutorBase): ...@@ -236,9 +240,11 @@ class TestMNIST(TestParallelExecutorBase):
def test_simple_fc(self): def test_simple_fc(self):
self.check_network_convergence(simple_fc_net) self.check_network_convergence(simple_fc_net)
self.check_network_convergence(simple_fc_net, allow_op_delay=True)
def test_batchnorm_fc(self): def test_batchnorm_fc(self):
self.check_network_convergence(fc_with_batchnorm) self.check_network_convergence(fc_with_batchnorm)
self.check_network_convergence(fc_with_batchnorm, allow_op_delay=True)
class TestResnet(TestParallelExecutorBase): class TestResnet(TestParallelExecutorBase):
...@@ -262,10 +268,10 @@ class TestResnet(TestParallelExecutorBase): ...@@ -262,10 +268,10 @@ class TestResnet(TestParallelExecutorBase):
def test_resnet(self): def test_resnet(self):
import functools import functools
batch_size = 4 batch_size = 2
self.check_network_convergence( self.check_network_convergence(
functools.partial( functools.partial(
SE_ResNeXt152, batch_size=batch_size), SE_ResNeXt152Small, batch_size=batch_size),
iter=20, iter=20,
batch_size=batch_size) batch_size=batch_size)
......
python/paddle/fluid/tests/unittests/test_recv_op.py
浏览文件 @ 1f6e0448
...@@ -23,7 +23,7 @@ import time ...@@ -23,7 +23,7 @@ import time
class TestRecvOp(unittest.TestCase): class TestRecvOp(unittest.TestCase):
def test_send(self): def no_test_send(self):
# Run init_serv in a thread # Run init_serv in a thread
place = fluid.CPUPlace() place = fluid.CPUPlace()
p = Process(target=self.init_serv, args=(place, )) p = Process(target=self.init_serv, args=(place, ))
......
python/paddle/fluid/tests/unittests/test_reshape_op.py
浏览文件 @ 1f6e0448
...@@ -14,15 +14,19 @@ ...@@ -14,15 +14,19 @@
import unittest import unittest
import numpy as np import numpy as np
from op_test import OpTest from op_test import OpTest
class TestReshapeOp(OpTest): class TestReshapeOp(OpTest):
def setUp(self): def setUp(self):
ori_shape = (2, 25)
new_shape = (5, 10)
self.op_type = "reshape" self.op_type = "reshape"
self.inputs = {'X': np.random.random((10, 20)).astype("float32")} self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {'shape': [10 * 20]} self.attrs = {"shape": new_shape, "inplace": False}
self.outputs = {'Out': self.inputs['X'].reshape(self.attrs['shape'])} self.outputs = {"Out": self.inputs["X"].reshape(new_shape)}
def test_check_output(self): def test_check_output(self):
self.check_output() self.check_output()
...@@ -31,12 +35,33 @@ class TestReshapeOp(OpTest): ...@@ -31,12 +35,33 @@ class TestReshapeOp(OpTest):
self.check_grad(["X"], "Out") self.check_grad(["X"], "Out")
class TestReshapeOpDimInfer(OpTest): class TestReshapeOpDimInfer1(OpTest):
def setUp(self): def setUp(self):
ori_shape = (5, 10)
new_shape = (5, -1, 5)
self.op_type = "reshape" self.op_type = "reshape"
self.inputs = {'X': np.random.random((10, 20)).astype("float32")} self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {'shape': [4, -1, 5]} self.attrs = {"shape": new_shape, "inplace": False}
self.outputs = {'Out': self.inputs['X'].reshape(self.attrs['shape'])} self.outputs = {"Out": self.inputs["X"].reshape(self.attrs["shape"])}
def test_check_output(self):
self.check_output()
def test_check_grad(self):
self.check_grad(["X"], "Out")
class TestReshapeOpDimInfer2(OpTest):
def setUp(self):
ori_shape = (2, 2, 6)
new_shape = (2, 0, 3, -1)
infered_shape = (2, 2, 3, -1)
self.op_type = "reshape"
self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {"shape": new_shape, "inplace": False}
self.outputs = {"Out": self.inputs["X"].reshape(infered_shape)}
def test_check_output(self): def test_check_output(self):
self.check_output() self.check_output()
...@@ -47,10 +72,30 @@ class TestReshapeOpDimInfer(OpTest): ...@@ -47,10 +72,30 @@ class TestReshapeOpDimInfer(OpTest):
class TestReshapeOpInplace(OpTest): class TestReshapeOpInplace(OpTest):
def setUp(self): def setUp(self):
ori_shape = (2, 25)
new_shape = (5, 10)
self.op_type = "reshape"
self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {"shape": new_shape}
self.outputs = {"Out": self.inputs["X"].reshape(new_shape)}
def test_check_output(self):
self.check_output()
def test_check_grad(self):
self.check_grad(["X"], "Out")
class TestReshapeOpDimInferInplace1(OpTest):
def setUp(self):
ori_shape = (5, 10)
new_shape = (5, -1, 5)
self.op_type = "reshape" self.op_type = "reshape"
self.inputs = {'X': np.random.random((10, 20)).astype("float32")} self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {'shape': [10 * 20], 'inplace': True} self.attrs = {"shape": new_shape}
self.outputs = {'Out': self.inputs['X'].reshape(self.attrs['shape'])} self.outputs = {"Out": self.inputs["X"].reshape(new_shape)}
def test_check_output(self): def test_check_output(self):
self.check_output() self.check_output()
...@@ -59,12 +104,38 @@ class TestReshapeOpInplace(OpTest): ...@@ -59,12 +104,38 @@ class TestReshapeOpInplace(OpTest):
self.check_grad(["X"], "Out") self.check_grad(["X"], "Out")
class TestReshapeOpDimInferInplace(OpTest): class TestReshapeOpDimInferInplace2(OpTest):
def setUp(self): def setUp(self):
ori_shape = (2, 2, 6)
new_shape = (2, 0, 3, -1)
infered_shape = (2, 2, 3, -1)
self.op_type = "reshape"
self.inputs = {"X": np.random.random(ori_shape).astype("float32")}
self.attrs = {"shape": new_shape}
self.outputs = {"Out": self.inputs["X"].reshape(infered_shape)}
def test_check_output(self):
self.check_output()
def test_check_grad(self):
self.check_grad(["X"], "Out")
class TestReshapeOpWithInputShape(OpTest):
def setUp(self):
ori_shape = (6, 5)
new_shape = (0, -1, 5)
actual_shape = (2, 3, 5)
self.op_type = "reshape" self.op_type = "reshape"
self.inputs = {'X': np.random.random((10, 20)).astype("float32")} self.inputs = {
self.attrs = {'shape': [4, -1, 5], 'inplace': True} "X": np.random.random(ori_shape).astype("float32"),
self.outputs = {'Out': self.inputs['X'].reshape(self.attrs['shape'])} "Shape": np.array(
actual_shape, dtype="int32")
}
self.attrs = {"shape": new_shape}
self.outputs = {"Out": self.inputs["X"].reshape(actual_shape)}
def test_check_output(self): def test_check_output(self):
self.check_output() self.check_output()
...@@ -73,5 +144,5 @@ class TestReshapeOpDimInferInplace(OpTest): ...@@ -73,5 +144,5 @@ class TestReshapeOpDimInferInplace(OpTest):
self.check_grad(["X"], "Out") self.check_grad(["X"], "Out")
if __name__ == '__main__': if __name__ == "__main__":
unittest.main() unittest.main()
python/paddle/fluid/tests/unittests/test_target_assign_op.py 100755 → 100644
浏览文件 @ 1f6e0448
文件模式从 100755 更改为 100644
python/paddle/v2/__init__.py
浏览文件 @ 1f6e0448
...@@ -22,13 +22,17 @@ import data_type ...@@ -22,13 +22,17 @@ import data_type
import topology import topology
import networks import networks
import evaluator import evaluator
from . import dataset
from . import reader
from . import plot from . import plot
import attr import attr
import op import op
import pooling import pooling
import inference import inference
import networks import networks
import minibatch
import plot import plot
import image
import paddle.trainer.config_parser as cp import paddle.trainer.config_parser as cp
__all__ = [ __all__ = [
...@@ -44,11 +48,14 @@ __all__ = [ ...@@ -44,11 +48,14 @@ __all__ = [
'data_type', 'data_type',
'attr', 'attr',
'pooling', 'pooling',
'dataset',
'reader',
'topology', 'topology',
'networks', 'networks',
'infer', 'infer',
'plot', 'plot',
'evaluator', 'evaluator',
'image',
'master', 'master',
] ]
...@@ -146,3 +153,4 @@ def init(**kwargs): ...@@ -146,3 +153,4 @@ def init(**kwargs):
infer = inference.infer infer = inference.infer
batch = minibatch.batch
python/paddle/v2/dataset/__init__.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Dataset package.
"""
import mnist
import imikolov
import imdb
import cifar
import movielens
import conll05
import uci_housing
import sentiment
import wmt14
import wmt16
import mq2007
import flowers
import voc2012
__all__ = [
'mnist',
'imikolov',
'imdb',
'cifar',
'movielens',
'conll05',
'sentiment',
'uci_housing',
'wmt14',
'wmt16',
'mq2007',
'flowers',
'voc2012',
]
python/paddle/v2/dataset/cifar.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
CIFAR dataset.
This module will download dataset from
https://www.cs.toronto.edu/~kriz/cifar.html and parse train/test set into
paddle reader creators.
The CIFAR-10 dataset consists of 60000 32x32 colour images in 10 classes,
with 6000 images per class. There are 50000 training images and 10000 test
images.
The CIFAR-100 dataset is just like the CIFAR-10, except it has 100 classes
containing 600 images each. There are 500 training images and 100 testing
images per class.
"""
import cPickle
import itertools
import numpy
import paddle.v2.dataset.common
import tarfile
__all__ = ['train100', 'test100', 'train10', 'test10', 'convert']
URL_PREFIX = 'https://www.cs.toronto.edu/~kriz/'
CIFAR10_URL = URL_PREFIX + 'cifar-10-python.tar.gz'
CIFAR10_MD5 = 'c58f30108f718f92721af3b95e74349a'
CIFAR100_URL = URL_PREFIX + 'cifar-100-python.tar.gz'
CIFAR100_MD5 = 'eb9058c3a382ffc7106e4002c42a8d85'
def reader_creator(filename, sub_name):
def read_batch(batch):
data = batch['data']
labels = batch.get('labels', batch.get('fine_labels', None))
assert labels is not None
for sample, label in itertools.izip(data, labels):
yield (sample / 255.0).astype(numpy.float32), int(label)
def reader():
with tarfile.open(filename, mode='r') as f:
names = (each_item.name for each_item in f
if sub_name in each_item.name)
for name in names:
batch = cPickle.load(f.extractfile(name))
for item in read_batch(batch):
yield item
return reader
def train100():
"""
CIFAR-100 training set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 99].
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(CIFAR100_URL, 'cifar', CIFAR100_MD5),
'train')
def test100():
"""
CIFAR-100 test set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Test reader creator.
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(CIFAR100_URL, 'cifar', CIFAR100_MD5),
'test')
def train10():
"""
CIFAR-10 training set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(CIFAR10_URL, 'cifar', CIFAR10_MD5),
'data_batch')
def test10():
"""
CIFAR-10 test set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Test reader creator.
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(CIFAR10_URL, 'cifar', CIFAR10_MD5),
'test_batch')
def fetch():
paddle.v2.dataset.common.download(CIFAR10_URL, 'cifar', CIFAR10_MD5)
paddle.v2.dataset.common.download(CIFAR100_URL, 'cifar', CIFAR100_MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, train100(), 1000, "cifar_train100")
paddle.v2.dataset.common.convert(path, test100(), 1000, "cifar_test100")
paddle.v2.dataset.common.convert(path, train10(), 1000, "cifar_train10")
paddle.v2.dataset.common.convert(path, test10(), 1000, "cifar_test10")
python/paddle/v2/dataset/common.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import requests
import hashlib
import os
import errno
import shutil
import sys
import importlib
import paddle.v2.dataset
import cPickle
import glob
import cPickle as pickle
__all__ = [
'DATA_HOME',
'download',
'md5file',
'split',
'cluster_files_reader',
'convert',
]
DATA_HOME = os.path.expanduser('~/.cache/paddle/dataset')
# When running unit tests, there could be multiple processes that
# trying to create DATA_HOME directory simultaneously, so we cannot
# use a if condition to check for the existence of the directory;
# instead, we use the filesystem as the synchronization mechanism by
# catching returned errors.
def must_mkdirs(path):
try:
os.makedirs(DATA_HOME)
except OSError as exc:
if exc.errno != errno.EEXIST:
raise
pass
must_mkdirs(DATA_HOME)
def md5file(fname):
hash_md5 = hashlib.md5()
f = open(fname, "rb")
for chunk in iter(lambda: f.read(4096), b""):
hash_md5.update(chunk)
f.close()
return hash_md5.hexdigest()
def download(url, module_name, md5sum, save_name=None):
dirname = os.path.join(DATA_HOME, module_name)
if not os.path.exists(dirname):
os.makedirs(dirname)
filename = os.path.join(dirname,
url.split('/')[-1]
if save_name is None else save_name)
retry = 0
retry_limit = 3
while not (os.path.exists(filename) and md5file(filename) == md5sum):
if os.path.exists(filename):
print "file md5", md5file(filename), md5sum
if retry < retry_limit:
retry += 1
else:
raise RuntimeError("Cannot download {0} within retry limit {1}".
format(url, retry_limit))
print "Cache file %s not found, downloading %s" % (filename, url)
r = requests.get(url, stream=True)
total_length = r.headers.get('content-length')
if total_length is None:
with open(filename, 'w') as f:
shutil.copyfileobj(r.raw, f)
else:
with open(filename, 'w') as f:
dl = 0
total_length = int(total_length)
for data in r.iter_content(chunk_size=4096):
dl += len(data)
f.write(data)
done = int(50 * dl / total_length)
sys.stdout.write("\r[%s%s]" % ('=' * done,
' ' * (50 - done)))
sys.stdout.flush()
return filename
def fetch_all():
for module_name in filter(lambda x: not x.startswith("__"),
dir(paddle.v2.dataset)):
if "fetch" in dir(
importlib.import_module("paddle.v2.dataset.%s" % module_name)):
getattr(
importlib.import_module("paddle.v2.dataset.%s" % module_name),
"fetch")()
def fetch_all_recordio(path):
for module_name in filter(lambda x: not x.startswith("__"),
dir(paddle.v2.dataset)):
if "convert" in dir(
importlib.import_module("paddle.v2.dataset.%s" % module_name)) and \
not module_name == "common":
ds_path = os.path.join(path, module_name)
must_mkdirs(ds_path)
getattr(
importlib.import_module("paddle.v2.dataset.%s" % module_name),
"convert")(ds_path)
def split(reader, line_count, suffix="%05d.pickle", dumper=cPickle.dump):
"""
you can call the function as:
split(paddle.v2.dataset.cifar.train10(), line_count=1000,
suffix="imikolov-train-%05d.pickle")
the output files as:
|-imikolov-train-00000.pickle
|-imikolov-train-00001.pickle
|- ...
|-imikolov-train-00480.pickle
:param reader: is a reader creator
:param line_count: line count for each file
:param suffix: the suffix for the output files, should contain "%d"
means the id for each file. Default is "%05d.pickle"
:param dumper: is a callable function that dump object to file, this
function will be called as dumper(obj, f) and obj is the object
will be dumped, f is a file object. Default is cPickle.dump.
"""
if not callable(dumper):
raise TypeError("dumper should be callable.")
lines = []
indx_f = 0
for i, d in enumerate(reader()):
lines.append(d)
if i >= line_count and i % line_count == 0:
with open(suffix % indx_f, "w") as f:
dumper(lines, f)
lines = []
indx_f += 1
if lines:
with open(suffix % indx_f, "w") as f:
dumper(lines, f)
def cluster_files_reader(files_pattern,
trainer_count,
trainer_id,
loader=cPickle.load):
"""
Create a reader that yield element from the given files, select
a file set according trainer count and trainer_id
:param files_pattern: the files which generating by split(...)
:param trainer_count: total trainer count
:param trainer_id: the trainer rank id
:param loader: is a callable function that load object from file, this
function will be called as loader(f) and f is a file object.
Default is cPickle.load
"""
def reader():
if not callable(loader):
raise TypeError("loader should be callable.")
file_list = glob.glob(files_pattern)
file_list.sort()
my_file_list = []
for idx, fn in enumerate(file_list):
if idx % trainer_count == trainer_id:
print "append file: %s" % fn
my_file_list.append(fn)
for fn in my_file_list:
with open(fn, "r") as f:
lines = loader(f)
for line in lines:
yield line
return reader
def convert(output_path, reader, line_count, name_prefix):
import recordio
"""
Convert data from reader to recordio format files.
:param output_path: directory in which output files will be saved.
:param reader: a data reader, from which the convert program will read
data instances.
:param name_prefix: the name prefix of generated files.
:param max_lines_to_shuffle: the max lines numbers to shuffle before
writing.
"""
assert line_count >= 1
indx_f = 0
def write_data(indx_f, lines):
filename = "%s/%s-%05d" % (output_path, name_prefix, indx_f)
writer = recordio.writer(filename)
for l in lines:
# FIXME(Yancey1989):
# dumps with protocol: pickle.HIGHEST_PROTOCOL
writer.write(cPickle.dumps(l))
writer.close()
lines = []
for i, d in enumerate(reader()):
lines.append(d)
if i % line_count == 0 and i >= line_count:
write_data(indx_f, lines)
lines = []
indx_f += 1
continue
write_data(indx_f, lines)
python/paddle/v2/dataset/conll05.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Conll05 dataset.
Paddle semantic role labeling Book and demo use this dataset as an example.
Because Conll05 is not free in public, the default downloaded URL is test set
of Conll05 (which is public). Users can change URL and MD5 to their Conll
dataset. And a pre-trained word vector model based on Wikipedia corpus is used
to initialize SRL model.
"""
import tarfile
import gzip
import itertools
import paddle.v2.dataset.common
__all__ = ['test, get_dict', 'get_embedding', 'convert']
DATA_URL = 'http://www.cs.upc.edu/~srlconll/conll05st-tests.tar.gz'
DATA_MD5 = '387719152ae52d60422c016e92a742fc'
WORDDICT_URL = 'http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/wordDict.txt'
WORDDICT_MD5 = 'ea7fb7d4c75cc6254716f0177a506baa'
VERBDICT_URL = 'http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/verbDict.txt'
VERBDICT_MD5 = '0d2977293bbb6cbefab5b0f97db1e77c'
TRGDICT_URL = 'http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/targetDict.txt'
TRGDICT_MD5 = 'd8c7f03ceb5fc2e5a0fa7503a4353751'
EMB_URL = 'http://paddlepaddle.bj.bcebos.com/demo/srl_dict_and_embedding/emb'
EMB_MD5 = 'bf436eb0faa1f6f9103017f8be57cdb7'
UNK_IDX = 0
def load_label_dict(filename):
d = dict()
tag_dict = set()
with open(filename, 'r') as f:
for i, line in enumerate(f):
line = line.strip()
if line.startswith("B-"):
tag_dict.add(line[2:])
elif line.startswith("I-"):
tag_dict.add(line[2:])
index = 0
for tag in tag_dict:
d["B-" + tag] = index
index += 1
d["I-" + tag] = index
index += 1
d["O"] = index
return d
def load_dict(filename):
d = dict()
with open(filename, 'r') as f:
for i, line in enumerate(f):
d[line.strip()] = i
return d
def corpus_reader(data_path, words_name, props_name):
"""
Read one corpus. It returns an iterator. Each element of
this iterator is a tuple including sentence and labels. The sentence is
consist of a list of word IDs. The labels include a list of label IDs.
:return: a iterator of data.
:rtype: iterator
"""
def reader():
tf = tarfile.open(data_path)
wf = tf.extractfile(words_name)
pf = tf.extractfile(props_name)
with gzip.GzipFile(fileobj=wf) as words_file, gzip.GzipFile(
fileobj=pf) as props_file:
sentences = []
labels = []
one_seg = []
for word, label in itertools.izip(words_file, props_file):
word = word.strip()
label = label.strip().split()
if len(label) == 0: # end of sentence
for i in xrange(len(one_seg[0])):
a_kind_lable = [x[i] for x in one_seg]
labels.append(a_kind_lable)
if len(labels) >= 1:
verb_list = []
for x in labels[0]:
if x != '-':
verb_list.append(x)
for i, lbl in enumerate(labels[1:]):
cur_tag = 'O'
is_in_bracket = False
lbl_seq = []
verb_word = ''
for l in lbl:
if l == '*' and is_in_bracket == False:
lbl_seq.append('O')
elif l == '*' and is_in_bracket == True:
lbl_seq.append('I-' + cur_tag)
elif l == '*)':
lbl_seq.append('I-' + cur_tag)
is_in_bracket = False
elif l.find('(') != -1 and l.find(')') != -1:
cur_tag = l[1:l.find('*')]
lbl_seq.append('B-' + cur_tag)
is_in_bracket = False
elif l.find('(') != -1 and l.find(')') == -1:
cur_tag = l[1:l.find('*')]
lbl_seq.append('B-' + cur_tag)
is_in_bracket = True
else:
raise RuntimeError('Unexpected label: %s' %
l)
yield sentences, verb_list[i], lbl_seq
sentences = []
labels = []
one_seg = []
else:
sentences.append(word)
one_seg.append(label)
pf.close()
wf.close()
tf.close()
return reader
def reader_creator(corpus_reader,
word_dict=None,
predicate_dict=None,
label_dict=None):
def reader():
for sentence, predicate, labels in corpus_reader():
sen_len = len(sentence)
verb_index = labels.index('B-V')
mark = [0] * len(labels)
if verb_index > 0:
mark[verb_index - 1] = 1
ctx_n1 = sentence[verb_index - 1]
else:
ctx_n1 = 'bos'
if verb_index > 1:
mark[verb_index - 2] = 1
ctx_n2 = sentence[verb_index - 2]
else:
ctx_n2 = 'bos'
mark[verb_index] = 1
ctx_0 = sentence[verb_index]
if verb_index < len(labels) - 1:
mark[verb_index + 1] = 1
ctx_p1 = sentence[verb_index + 1]
else:
ctx_p1 = 'eos'
if verb_index < len(labels) - 2:
mark[verb_index + 2] = 1
ctx_p2 = sentence[verb_index + 2]
else:
ctx_p2 = 'eos'
word_idx = [word_dict.get(w, UNK_IDX) for w in sentence]
ctx_n2_idx = [word_dict.get(ctx_n2, UNK_IDX)] * sen_len
ctx_n1_idx = [word_dict.get(ctx_n1, UNK_IDX)] * sen_len
ctx_0_idx = [word_dict.get(ctx_0, UNK_IDX)] * sen_len
ctx_p1_idx = [word_dict.get(ctx_p1, UNK_IDX)] * sen_len
ctx_p2_idx = [word_dict.get(ctx_p2, UNK_IDX)] * sen_len
pred_idx = [predicate_dict.get(predicate)] * sen_len
label_idx = [label_dict.get(w) for w in labels]
yield word_idx, ctx_n2_idx, ctx_n1_idx, \
ctx_0_idx, ctx_p1_idx, ctx_p2_idx, pred_idx, mark, label_idx
return reader
def get_dict():
"""
Get the word, verb and label dictionary of Wikipedia corpus.
"""
word_dict = load_dict(
paddle.v2.dataset.common.download(WORDDICT_URL, 'conll05st',
WORDDICT_MD5))
verb_dict = load_dict(
paddle.v2.dataset.common.download(VERBDICT_URL, 'conll05st',
VERBDICT_MD5))
label_dict = load_label_dict(
paddle.v2.dataset.common.download(TRGDICT_URL, 'conll05st',
TRGDICT_MD5))
return word_dict, verb_dict, label_dict
def get_embedding():
"""
Get the trained word vector based on Wikipedia corpus.
"""
return paddle.v2.dataset.common.download(EMB_URL, 'conll05st', EMB_MD5)
def test():
"""
Conll05 test set creator.
Because the training dataset is not free, the test dataset is used for
training. It returns a reader creator, each sample in the reader is nine
features, including sentence sequence, predicate, predicate context,
predicate context flag and tagged sequence.
:return: Training reader creator
:rtype: callable
"""
word_dict, verb_dict, label_dict = get_dict()
reader = corpus_reader(
paddle.v2.dataset.common.download(DATA_URL, 'conll05st', DATA_MD5),
words_name='conll05st-release/test.wsj/words/test.wsj.words.gz',
props_name='conll05st-release/test.wsj/props/test.wsj.props.gz')
return reader_creator(reader, word_dict, verb_dict, label_dict)
def fetch():
paddle.v2.dataset.common.download(WORDDICT_URL, 'conll05st', WORDDICT_MD5)
paddle.v2.dataset.common.download(VERBDICT_URL, 'conll05st', VERBDICT_MD5)
paddle.v2.dataset.common.download(TRGDICT_URL, 'conll05st', TRGDICT_MD5)
paddle.v2.dataset.common.download(EMB_URL, 'conll05st', EMB_MD5)
paddle.v2.dataset.common.download(DATA_URL, 'conll05st', DATA_MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, test(), 1000, "conl105_train")
paddle.v2.dataset.common.convert(path, test(), 1000, "conl105_test")
python/paddle/v2/dataset/flowers.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This module will download dataset from
http://www.robots.ox.ac.uk/~vgg/data/flowers/102/index.html
and parse train/test set intopaddle reader creators.
This set contains images of flowers belonging to 102 different categories.
The images were acquired by searching the web and taking pictures. There are a
minimum of 40 images for each category.
The database was used in:
Nilsback, M-E. and Zisserman, A. Automated flower classification over a large
number of classes.Proceedings of the Indian Conference on Computer Vision,
Graphics and Image Processing (2008)
http://www.robots.ox.ac.uk/~vgg/publications/papers/nilsback08.{pdf,ps.gz}.
"""
import cPickle
import itertools
import functools
from common import download
import tarfile
import scipy.io as scio
from paddle.v2.image import *
from paddle.v2.reader import *
import os
import numpy as np
from multiprocessing import cpu_count
__all__ = ['train', 'test', 'valid']
DATA_URL = 'http://www.robots.ox.ac.uk/~vgg/data/flowers/102/102flowers.tgz'
LABEL_URL = 'http://www.robots.ox.ac.uk/~vgg/data/flowers/102/imagelabels.mat'
SETID_URL = 'http://www.robots.ox.ac.uk/~vgg/data/flowers/102/setid.mat'
DATA_MD5 = '33bfc11892f1e405ca193ae9a9f2a118'
LABEL_MD5 = 'e0620be6f572b9609742df49c70aed4d'
SETID_MD5 = 'a5357ecc9cb78c4bef273ce3793fc85c'
# In official 'readme', tstid is the flag of test data
# and trnid is the flag of train data. But test data is more than train data.
# So we exchange the train data and test data.
TRAIN_FLAG = 'tstid'
TEST_FLAG = 'trnid'
VALID_FLAG = 'valid'
def default_mapper(is_train, sample):
'''
map image bytes data to type needed by model input layer
'''
img, label = sample
img = load_image_bytes(img)
img = simple_transform(
img, 256, 224, is_train, mean=[103.94, 116.78, 123.68])
return img.flatten().astype('float32'), label
train_mapper = functools.partial(default_mapper, True)
test_mapper = functools.partial(default_mapper, False)
def reader_creator(data_file,
label_file,
setid_file,
dataset_name,
mapper,
buffered_size=1024,
use_xmap=True):
'''
1. read images from tar file and
merge images into batch files in 102flowers.tgz_batch/
2. get a reader to read sample from batch file
:param data_file: downloaded data file
:type data_file: string
:param label_file: downloaded label file
:type label_file: string
:param setid_file: downloaded setid file containing information
about how to split dataset
:type setid_file: string
:param dataset_name: data set name (tstid|trnid|valid)
:type dataset_name: string
:param mapper: a function to map image bytes data to type
needed by model input layer
:type mapper: callable
:param buffered_size: the size of buffer used to process images
:type buffered_size: int
:return: data reader
:rtype: callable
'''
labels = scio.loadmat(label_file)['labels'][0]
indexes = scio.loadmat(setid_file)[dataset_name][0]
img2label = {}
for i in indexes:
img = "jpg/image_%05d.jpg" % i
img2label[img] = labels[i - 1]
file_list = batch_images_from_tar(data_file, dataset_name, img2label)
def reader():
for file in open(file_list):
file = file.strip()
batch = None
with open(file, 'r') as f:
batch = cPickle.load(f)
data = batch['data']
labels = batch['label']
for sample, label in itertools.izip(data, batch['label']):
yield sample, int(label) - 1
if use_xmap:
return xmap_readers(mapper, reader, cpu_count(), buffered_size)
else:
return map_readers(mapper, reader)
def train(mapper=train_mapper, buffered_size=1024, use_xmap=True):
'''
Create flowers training set reader.
It returns a reader, each sample in the reader is
image pixels in [0, 1] and label in [1, 102]
translated from original color image by steps:
1. resize to 256*256
2. random crop to 224*224
3. flatten
:param mapper: a function to map sample.
:type mapper: callable
:param buffered_size: the size of buffer used to process images
:type buffered_size: int
:return: train data reader
:rtype: callable
'''
return reader_creator(
download(DATA_URL, 'flowers', DATA_MD5),
download(LABEL_URL, 'flowers', LABEL_MD5),
download(SETID_URL, 'flowers', SETID_MD5), TRAIN_FLAG, mapper,
buffered_size, use_xmap)
def test(mapper=test_mapper, buffered_size=1024, use_xmap=True):
'''
Create flowers test set reader.
It returns a reader, each sample in the reader is
image pixels in [0, 1] and label in [1, 102]
translated from original color image by steps:
1. resize to 256*256
2. random crop to 224*224
3. flatten
:param mapper: a function to map sample.
:type mapper: callable
:param buffered_size: the size of buffer used to process images
:type buffered_size: int
:return: test data reader
:rtype: callable
'''
return reader_creator(
download(DATA_URL, 'flowers', DATA_MD5),
download(LABEL_URL, 'flowers', LABEL_MD5),
download(SETID_URL, 'flowers', SETID_MD5), TEST_FLAG, mapper,
buffered_size, use_xmap)
def valid(mapper=test_mapper, buffered_size=1024, use_xmap=True):
'''
Create flowers validation set reader.
It returns a reader, each sample in the reader is
image pixels in [0, 1] and label in [1, 102]
translated from original color image by steps:
1. resize to 256*256
2. random crop to 224*224
3. flatten
:param mapper: a function to map sample.
:type mapper: callable
:param buffered_size: the size of buffer used to process images
:type buffered_size: int
:return: test data reader
:rtype: callable
'''
return reader_creator(
download(DATA_URL, 'flowers', DATA_MD5),
download(LABEL_URL, 'flowers', LABEL_MD5),
download(SETID_URL, 'flowers', SETID_MD5), VALID_FLAG, mapper,
buffered_size, use_xmap)
def fetch():
download(DATA_URL, 'flowers', DATA_MD5)
download(LABEL_URL, 'flowers', LABEL_MD5)
download(SETID_URL, 'flowers', SETID_MD5)
python/paddle/v2/dataset/imdb.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
IMDB dataset.
This module downloads IMDB dataset from
http://ai.stanford.edu/%7Eamaas/data/sentiment/. This dataset contains a set
of 25,000 highly polar movie reviews for training, and 25,000 for testing.
Besides, this module also provides API for building dictionary.
"""
import paddle.v2.dataset.common
import collections
import tarfile
import re
import string
__all__ = ['build_dict', 'train', 'test', 'convert']
URL = 'http://ai.stanford.edu/%7Eamaas/data/sentiment/aclImdb_v1.tar.gz'
MD5 = '7c2ac02c03563afcf9b574c7e56c153a'
def tokenize(pattern):
"""
Read files that match the given pattern. Tokenize and yield each file.
"""
with tarfile.open(paddle.v2.dataset.common.download(URL, 'imdb',
MD5)) as tarf:
# Note that we should use tarfile.next(), which does
# sequential access of member files, other than
# tarfile.extractfile, which does random access and might
# destroy hard disks.
tf = tarf.next()
while tf != None:
if bool(pattern.match(tf.name)):
# newline and punctuations removal and ad-hoc tokenization.
yield tarf.extractfile(tf).read().rstrip("\n\r").translate(
None, string.punctuation).lower().split()
tf = tarf.next()
def build_dict(pattern, cutoff):
"""
Build a word dictionary from the corpus. Keys of the dictionary are words,
and values are zero-based IDs of these words.
"""
word_freq = collections.defaultdict(int)
for doc in tokenize(pattern):
for word in doc:
word_freq[word] += 1
# Not sure if we should prune less-frequent words here.
word_freq = filter(lambda x: x[1] > cutoff, word_freq.items())
dictionary = sorted(word_freq, key=lambda x: (-x[1], x[0]))
words, _ = list(zip(*dictionary))
word_idx = dict(zip(words, xrange(len(words))))
word_idx['<unk>'] = len(words)
return word_idx
def reader_creator(pos_pattern, neg_pattern, word_idx):
UNK = word_idx['<unk>']
INS = []
def load(pattern, out, label):
for doc in tokenize(pattern):
out.append(([word_idx.get(w, UNK) for w in doc], label))
load(pos_pattern, INS, 0)
load(neg_pattern, INS, 1)
def reader():
for doc, label in INS:
yield doc, label
return reader
def train(word_idx):
"""
IMDB training set creator.
It returns a reader creator, each sample in the reader is an zero-based ID
sequence and label in [0, 1].
:param word_idx: word dictionary
:type word_idx: dict
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
re.compile("aclImdb/train/pos/.*\.txt$"),
re.compile("aclImdb/train/neg/.*\.txt$"), word_idx)
def test(word_idx):
"""
IMDB test set creator.
It returns a reader creator, each sample in the reader is an zero-based ID
sequence and label in [0, 1].
:param word_idx: word dictionary
:type word_idx: dict
:return: Test reader creator
:rtype: callable
"""
return reader_creator(
re.compile("aclImdb/test/pos/.*\.txt$"),
re.compile("aclImdb/test/neg/.*\.txt$"), word_idx)
def word_dict():
"""
Build a word dictionary from the corpus.
:return: Word dictionary
:rtype: dict
"""
return build_dict(
re.compile("aclImdb/((train)|(test))/((pos)|(neg))/.*\.txt$"), 150)
def fetch():
paddle.v2.dataset.common.download(URL, 'imdb', MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
w = word_dict()
paddle.v2.dataset.common.convert(path, lambda: train(w), 1000, "imdb_train")
paddle.v2.dataset.common.convert(path, lambda: test(w), 1000, "imdb_test")
python/paddle/v2/dataset/imikolov.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
imikolov's simple dataset.
This module will download dataset from
http://www.fit.vutbr.cz/~imikolov/rnnlm/ and parse training set and test set
into paddle reader creators.
"""
import paddle.v2.dataset.common
import collections
import tarfile
__all__ = ['train', 'test', 'build_dict', 'convert']
URL = 'http://www.fit.vutbr.cz/~imikolov/rnnlm/simple-examples.tgz'
MD5 = '30177ea32e27c525793142b6bf2c8e2d'
class DataType(object):
NGRAM = 1
SEQ = 2
def word_count(f, word_freq=None):
if word_freq is None:
word_freq = collections.defaultdict(int)
for l in f:
for w in l.strip().split():
word_freq[w] += 1
word_freq['<s>'] += 1
word_freq['<e>'] += 1
return word_freq
def build_dict(min_word_freq=50):
"""
Build a word dictionary from the corpus, Keys of the dictionary are words,
and values are zero-based IDs of these words.
"""
train_filename = './simple-examples/data/ptb.train.txt'
test_filename = './simple-examples/data/ptb.valid.txt'
with tarfile.open(
paddle.v2.dataset.common.download(
paddle.v2.dataset.imikolov.URL, 'imikolov',
paddle.v2.dataset.imikolov.MD5)) as tf:
trainf = tf.extractfile(train_filename)
testf = tf.extractfile(test_filename)
word_freq = word_count(testf, word_count(trainf))
if '<unk>' in word_freq:
# remove <unk> for now, since we will set it as last index
del word_freq['<unk>']
word_freq = filter(lambda x: x[1] > min_word_freq, word_freq.items())
word_freq_sorted = sorted(word_freq, key=lambda x: (-x[1], x[0]))
words, _ = list(zip(*word_freq_sorted))
word_idx = dict(zip(words, xrange(len(words))))
word_idx['<unk>'] = len(words)
return word_idx
def reader_creator(filename, word_idx, n, data_type):
def reader():
with tarfile.open(
paddle.v2.dataset.common.download(
paddle.v2.dataset.imikolov.URL, 'imikolov',
paddle.v2.dataset.imikolov.MD5)) as tf:
f = tf.extractfile(filename)
UNK = word_idx['<unk>']
for l in f:
if DataType.NGRAM == data_type:
assert n > -1, 'Invalid gram length'
l = ['<s>'] + l.strip().split() + ['<e>']
if len(l) >= n:
l = [word_idx.get(w, UNK) for w in l]
for i in range(n, len(l) + 1):
yield tuple(l[i - n:i])
elif DataType.SEQ == data_type:
l = l.strip().split()
l = [word_idx.get(w, UNK) for w in l]
src_seq = [word_idx['<s>']] + l
trg_seq = l + [word_idx['<e>']]
if n > 0 and len(src_seq) > n: continue
yield src_seq, trg_seq
else:
assert False, 'Unknow data type'
return reader
def train(word_idx, n, data_type=DataType.NGRAM):
"""
imikolov training set creator.
It returns a reader creator, each sample in the reader is a word ID
tuple.
:param word_idx: word dictionary
:type word_idx: dict
:param n: sliding window size if type is ngram, otherwise max length of sequence
:type n: int
:param data_type: data type (ngram or sequence)
:type data_type: member variable of DataType (NGRAM or SEQ)
:return: Training reader creator
:rtype: callable
"""
return reader_creator('./simple-examples/data/ptb.train.txt', word_idx, n,
data_type)
def test(word_idx, n, data_type=DataType.NGRAM):
"""
imikolov test set creator.
It returns a reader creator, each sample in the reader is a word ID
tuple.
:param word_idx: word dictionary
:type word_idx: dict
:param n: sliding window size if type is ngram, otherwise max length of sequence
:type n: int
:param data_type: data type (ngram or sequence)
:type data_type: member variable of DataType (NGRAM or SEQ)
:return: Test reader creator
:rtype: callable
"""
return reader_creator('./simple-examples/data/ptb.valid.txt', word_idx, n,
data_type)
def fetch():
paddle.v2.dataset.common.download(URL, "imikolov", MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
N = 5
word_dict = build_dict()
paddle.v2.dataset.common.convert(path,
train(word_dict, N), 1000,
"imikolov_train")
paddle.v2.dataset.common.convert(path,
test(word_dict, N), 1000, "imikolov_test")
python/paddle/v2/dataset/mnist.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
MNIST dataset.
This module will download dataset from http://yann.lecun.com/exdb/mnist/ and
parse training set and test set into paddle reader creators.
"""
import paddle.v2.dataset.common
import subprocess
import numpy
import platform
__all__ = ['train', 'test', 'convert']
URL_PREFIX = 'http://yann.lecun.com/exdb/mnist/'
TEST_IMAGE_URL = URL_PREFIX + 't10k-images-idx3-ubyte.gz'
TEST_IMAGE_MD5 = '9fb629c4189551a2d022fa330f9573f3'
TEST_LABEL_URL = URL_PREFIX + 't10k-labels-idx1-ubyte.gz'
TEST_LABEL_MD5 = 'ec29112dd5afa0611ce80d1b7f02629c'
TRAIN_IMAGE_URL = URL_PREFIX + 'train-images-idx3-ubyte.gz'
TRAIN_IMAGE_MD5 = 'f68b3c2dcbeaaa9fbdd348bbdeb94873'
TRAIN_LABEL_URL = URL_PREFIX + 'train-labels-idx1-ubyte.gz'
TRAIN_LABEL_MD5 = 'd53e105ee54ea40749a09fcbcd1e9432'
def reader_creator(image_filename, label_filename, buffer_size):
def reader():
if platform.system() == 'Darwin':
zcat_cmd = 'gzcat'
elif platform.system() == 'Linux':
zcat_cmd = 'zcat'
else:
raise NotImplementedError()
# According to http://stackoverflow.com/a/38061619/724872, we
# cannot use standard package gzip here.
m = subprocess.Popen([zcat_cmd, image_filename], stdout=subprocess.PIPE)
m.stdout.read(16) # skip some magic bytes
l = subprocess.Popen([zcat_cmd, label_filename], stdout=subprocess.PIPE)
l.stdout.read(8) # skip some magic bytes
try: # reader could be break.
while True:
labels = numpy.fromfile(
l.stdout, 'ubyte', count=buffer_size).astype("int")
if labels.size != buffer_size:
break # numpy.fromfile returns empty slice after EOF.
images = numpy.fromfile(
m.stdout, 'ubyte', count=buffer_size * 28 * 28).reshape(
(buffer_size, 28 * 28)).astype('float32')
images = images / 255.0 * 2.0 - 1.0
for i in xrange(buffer_size):
yield images[i, :], int(labels[i])
finally:
m.terminate()
l.terminate()
return reader
def train():
"""
MNIST training set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(TRAIN_IMAGE_URL, 'mnist',
TRAIN_IMAGE_MD5),
paddle.v2.dataset.common.download(TRAIN_LABEL_URL, 'mnist',
TRAIN_LABEL_MD5), 100)
def test():
"""
MNIST test set creator.
It returns a reader creator, each sample in the reader is image pixels in
[0, 1] and label in [0, 9].
:return: Test reader creator.
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(TEST_IMAGE_URL, 'mnist',
TEST_IMAGE_MD5),
paddle.v2.dataset.common.download(TEST_LABEL_URL, 'mnist',
TEST_LABEL_MD5), 100)
def fetch():
paddle.v2.dataset.common.download(TRAIN_IMAGE_URL, 'mnist', TRAIN_IMAGE_MD5)
paddle.v2.dataset.common.download(TRAIN_LABEL_URL, 'mnist', TRAIN_LABEL_MD5)
paddle.v2.dataset.common.download(TEST_IMAGE_URL, 'mnist', TEST_IMAGE_MD5)
paddle.v2.dataset.common.download(TEST_LABEL_URL, 'mnist', TRAIN_LABEL_MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, train(), 1000, "minist_train")
paddle.v2.dataset.common.convert(path, test(), 1000, "minist_test")
python/paddle/v2/dataset/movielens.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Movielens 1-M dataset.
Movielens 1-M dataset contains 1 million ratings from 6000 users on 4000
movies, which was collected by GroupLens Research. This module will download
Movielens 1-M dataset from
http://files.grouplens.org/datasets/movielens/ml-1m.zip and parse training
set and test set into paddle reader creators.
"""
import zipfile
import paddle.v2.dataset.common
import re
import random
import functools
__all__ = [
'train', 'test', 'get_movie_title_dict', 'max_movie_id', 'max_user_id',
'age_table', 'movie_categories', 'max_job_id', 'user_info', 'movie_info',
'convert'
]
age_table = [1, 18, 25, 35, 45, 50, 56]
URL = 'http://files.grouplens.org/datasets/movielens/ml-1m.zip'
MD5 = 'c4d9eecfca2ab87c1945afe126590906'
class MovieInfo(object):
"""
Movie id, title and categories information are stored in MovieInfo.
"""
def __init__(self, index, categories, title):
self.index = int(index)
self.categories = categories
self.title = title
def value(self):
"""
Get information from a movie.
"""
return [
self.index, [CATEGORIES_DICT[c] for c in self.categories],
[MOVIE_TITLE_DICT[w.lower()] for w in self.title.split()]
]
def __str__(self):
return "<MovieInfo id(%d), title(%s), categories(%s)>" % (
self.index, self.title, self.categories)
def __repr__(self):
return self.__str__()
class UserInfo(object):
"""
User id, gender, age, and job information are stored in UserInfo.
"""
def __init__(self, index, gender, age, job_id):
self.index = int(index)
self.is_male = gender == 'M'
self.age = age_table.index(int(age))
self.job_id = int(job_id)
def value(self):
"""
Get information from a user.
"""
return [self.index, 0 if self.is_male else 1, self.age, self.job_id]
def __str__(self):
return "<UserInfo id(%d), gender(%s), age(%d), job(%d)>" % (
self.index, "M"
if self.is_male else "F", age_table[self.age], self.job_id)
def __repr__(self):
return str(self)
MOVIE_INFO = None
MOVIE_TITLE_DICT = None
CATEGORIES_DICT = None
USER_INFO = None
def __initialize_meta_info__():
fn = paddle.v2.dataset.common.download(URL, "movielens", MD5)
global MOVIE_INFO
if MOVIE_INFO is None:
pattern = re.compile(r'^(.*)\((\d+)\)$')
with zipfile.ZipFile(file=fn) as package:
for info in package.infolist():
assert isinstance(info, zipfile.ZipInfo)
MOVIE_INFO = dict()
title_word_set = set()
categories_set = set()
with package.open('ml-1m/movies.dat') as movie_file:
for i, line in enumerate(movie_file):
movie_id, title, categories = line.strip().split('::')
categories = categories.split('|')
for c in categories:
categories_set.add(c)
title = pattern.match(title).group(1)
MOVIE_INFO[int(movie_id)] = MovieInfo(
index=movie_id, categories=categories, title=title)
for w in title.split():
title_word_set.add(w.lower())
global MOVIE_TITLE_DICT
MOVIE_TITLE_DICT = dict()
for i, w in enumerate(title_word_set):
MOVIE_TITLE_DICT[w] = i
global CATEGORIES_DICT
CATEGORIES_DICT = dict()
for i, c in enumerate(categories_set):
CATEGORIES_DICT[c] = i
global USER_INFO
USER_INFO = dict()
with package.open('ml-1m/users.dat') as user_file:
for line in user_file:
uid, gender, age, job, _ = line.strip().split("::")
USER_INFO[int(uid)] = UserInfo(
index=uid, gender=gender, age=age, job_id=job)
return fn
def __reader__(rand_seed=0, test_ratio=0.1, is_test=False):
fn = __initialize_meta_info__()
rand = random.Random(x=rand_seed)
with zipfile.ZipFile(file=fn) as package:
with package.open('ml-1m/ratings.dat') as rating:
for line in rating:
if (rand.random() < test_ratio) == is_test:
uid, mov_id, rating, _ = line.strip().split("::")
uid = int(uid)
mov_id = int(mov_id)
rating = float(rating) * 2 - 5.0
mov = MOVIE_INFO[mov_id]
usr = USER_INFO[uid]
yield usr.value() + mov.value() + [[rating]]
def __reader_creator__(**kwargs):
return lambda: __reader__(**kwargs)
train = functools.partial(__reader_creator__, is_test=False)
test = functools.partial(__reader_creator__, is_test=True)
def get_movie_title_dict():
"""
Get movie title dictionary.
"""
__initialize_meta_info__()
return MOVIE_TITLE_DICT
def __max_index_info__(a, b):
if a.index > b.index:
return a
else:
return b
def max_movie_id():
"""
Get the maximum value of movie id.
"""
__initialize_meta_info__()
return reduce(__max_index_info__, MOVIE_INFO.viewvalues()).index
def max_user_id():
"""
Get the maximum value of user id.
"""
__initialize_meta_info__()
return reduce(__max_index_info__, USER_INFO.viewvalues()).index
def __max_job_id_impl__(a, b):
if a.job_id > b.job_id:
return a
else:
return b
def max_job_id():
"""
Get the maximum value of job id.
"""
__initialize_meta_info__()
return reduce(__max_job_id_impl__, USER_INFO.viewvalues()).job_id
def movie_categories():
"""
Get movie categoriges dictionary.
"""
__initialize_meta_info__()
return CATEGORIES_DICT
def user_info():
"""
Get user info dictionary.
"""
__initialize_meta_info__()
return USER_INFO
def movie_info():
"""
Get movie info dictionary.
"""
__initialize_meta_info__()
return MOVIE_INFO
def unittest():
for train_count, _ in enumerate(train()()):
pass
for test_count, _ in enumerate(test()()):
pass
print train_count, test_count
def fetch():
paddle.v2.dataset.common.download(URL, "movielens", MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, train(), 1000, "movielens_train")
paddle.v2.dataset.common.convert(path, test(), 1000, "movielens_test")
if __name__ == '__main__':
unittest()
python/paddle/v2/dataset/mq2007.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
MQ2007 dataset
MQ2007 is a query set from Million Query track of TREC 2007. There are about 1700 queries in it with labeled documents. In MQ2007, the 5-fold cross
validation strategy is adopted and the 5-fold partitions are included in the package. In each fold, there are three subsets for learning: training set,
validation set and testing set.
MQ2007 dataset from website
http://research.microsoft.com/en-us/um/beijing/projects/letor/LETOR4.0/Data/MQ2007.rar and parse training set and test set into paddle reader creators
"""
import os
import functools
import rarfile
from common import download
import numpy as np
# URL = "http://research.microsoft.com/en-us/um/beijing/projects/letor/LETOR4.0/Data/MQ2007.rar"
URL = "http://www.bigdatalab.ac.cn/benchmark/upload/download_source/7b6dbbe2-842c-11e4-a536-bcaec51b9163_MQ2007.rar"
MD5 = "7be1640ae95c6408dab0ae7207bdc706"
def __initialize_meta_info__():
"""
download and extract the MQ2007 dataset
"""
fn = fetch()
rar = rarfile.RarFile(fn)
dirpath = os.path.dirname(fn)
rar.extractall(path=dirpath)
return dirpath
class Query(object):
"""
queries used for learning to rank algorithms. It is created from relevance scores, query-document feature vectors
Parameters:
----------
query_id : int
query_id in dataset, mapping from query to relevance documents
relevance_score : int
relevance score of query and document pair
feature_vector : array, dense feature
feature in vector format
description : string
comment section in query doc pair data
"""
def __init__(self,
query_id=-1,
relevance_score=-1,
feature_vector=None,
description=""):
self.query_id = query_id
self.relevance_score = relevance_score
if feature_vector is None:
self.feature_vector = []
else:
self.feature_vector = feature_vector
self.description = description
def __str__(self):
string = "%s %s %s" % (str(self.relevance_score), str(self.query_id),
" ".join(str(f) for f in self.feature_vector))
return string
# @classmethod
def _parse_(self, text):
"""
parse line into Query
"""
comment_position = text.find('#')
line = text[:comment_position].strip()
self.description = text[comment_position + 1:].strip()
parts = line.split()
if len(parts) != 48:
sys.stdout.write("expect 48 space split parts, get %d" %
(len(parts)))
return None
# format : 0 qid:10 1:0.000272 2:0.000000 ....
self.relevance_score = int(parts[0])
self.query_id = int(parts[1].split(':')[1])
for p in parts[2:]:
pair = p.split(':')
self.feature_vector.append(float(pair[1]))
return self
class QueryList(object):
"""
group query into list, every item in list is a Query
"""
def __init__(self, querylist=None):
self.query_id = -1
if querylist is None:
self.querylist = []
else:
self.querylist = querylist
for query in self.querylist:
if self.query_id == -1:
self.query_id = query.query_id
else:
if self.query_id != query.query_id:
raise ValueError("query in list must be same query_id")
def __iter__(self):
for query in self.querylist:
yield query
def __len__(self):
return len(self.querylist)
def __getitem__(self, i):
return self.querylist[i]
def _correct_ranking_(self):
if self.querylist is None:
return
self.querylist.sort(key=lambda x: x.relevance_score, reverse=True)
def _add_query(self, query):
if self.query_id == -1:
self.query_id = query.query_id
else:
if self.query_id != query.query_id:
raise ValueError("query in list must be same query_id")
self.querylist.append(query)
def gen_plain_txt(querylist):
"""
gen plain text in list for other usage
Paramters:
--------
querylist : querylist, one query match many docment pairs in list, see QueryList
return :
------
query_id : np.array, shape=(samples_num, )
label : np.array, shape=(samples_num, )
querylist : np.array, shape=(samples_num, feature_dimension)
"""
if not isinstance(querylist, QueryList):
querylist = QueryList(querylist)
querylist._correct_ranking_()
for query in querylist:
yield querylist.query_id, query.relevance_score, np.array(
query.feature_vector)
def gen_point(querylist):
"""
gen item in list for point-wise learning to rank algorithm
Paramters:
--------
querylist : querylist, one query match many docment pairs in list, see QueryList
return :
------
label : np.array, shape=(samples_num, )
querylist : np.array, shape=(samples_num, feature_dimension)
"""
if not isinstance(querylist, QueryList):
querylist = QueryList(querylist)
querylist._correct_ranking_()
for query in querylist:
yield query.relevance_score, np.array(query.feature_vector)
def gen_pair(querylist, partial_order="full"):
"""
gen pair for pair-wise learning to rank algorithm
Paramters:
--------
querylist : querylist, one query match many docment pairs in list, see QueryList
pairtial_order : "full" or "neighbour"
there is redudant in all possiable pair combinations, which can be simplifed
gen pairs for neighbour items or the full partial order pairs
return :
------
label : np.array, shape=(1)
query_left : np.array, shape=(1, feature_dimension)
query_right : same as left
"""
if not isinstance(querylist, QueryList):
querylist = QueryList(querylist)
querylist._correct_ranking_()
labels = []
docpairs = []
# C(n,2)
for i in range(len(querylist)):
query_left = querylist[i]
for j in range(i + 1, len(querylist)):
query_right = querylist[j]
if query_left.relevance_score > query_right.relevance_score:
labels.append([1])
docpairs.append([
np.array(query_left.feature_vector),
np.array(query_right.feature_vector)
])
elif query_left.relevance_score < query_right.relevance_score:
labels.append([1])
docpairs.append([
np.array(query_right.feature_vector),
np.array(query_left.feature_vector)
])
for label, pair in zip(labels, docpairs):
yield np.array(label), pair[0], pair[1]
def gen_list(querylist):
"""
gen item in list for list-wise learning to rank algorithm
Paramters:
--------
querylist : querylist, one query match many docment pairs in list, see QueryList
return :
------
label : np.array, shape=(samples_num, )
querylist : np.array, shape=(samples_num, feature_dimension)
"""
if not isinstance(querylist, QueryList):
querylist = QueryList(querylist)
querylist._correct_ranking_()
relevance_score_list = [[query.relevance_score] for query in querylist]
feature_vector_list = [query.feature_vector for query in querylist]
yield np.array(relevance_score_list), np.array(feature_vector_list)
def query_filter(querylists):
"""
filter query get only document with label 0.
label 0, 1, 2 means the relevance score document with query
parameters :
querylist : QueyList list
return :
querylist : QueyList list
"""
filter_query = []
for querylist in querylists:
relevance_score_list = [query.relevance_score for query in querylist]
if sum(relevance_score_list) != .0:
filter_query.append(querylist)
return filter_query
def load_from_text(filepath, shuffle=False, fill_missing=-1):
"""
parse data file into querys
"""
prev_query_id = -1
querylists = []
querylist = None
fn = __initialize_meta_info__()
with open(os.path.join(fn, filepath)) as f:
for line in f:
query = Query()
query = query._parse_(line)
if query == None:
continue
if query.query_id != prev_query_id:
if querylist is not None:
querylists.append(querylist)
querylist = QueryList()
prev_query_id = query.query_id
querylist._add_query(query)
if querylist is not None:
querylists.append(querylist)
return querylists
def __reader__(filepath, format="pairwise", shuffle=False, fill_missing=-1):
"""
Parameters
--------
filename : string
fill_missing : fill the missing value. default in MQ2007 is -1
Returns
------
yield
label query_left, query_right # format = "pairwise"
label querylist # format = "listwise"
"""
querylists = query_filter(
load_from_text(
filepath, shuffle=shuffle, fill_missing=fill_missing))
for querylist in querylists:
if format == "plain_txt":
yield next(gen_plain_txt(querylist))
elif format == "pointwise":
yield next(gen_point(querylist))
elif format == "pairwise":
for pair in gen_pair(querylist):
yield pair
elif format == "listwise":
yield next(gen_list(querylist))
train = functools.partial(__reader__, filepath="MQ2007/MQ2007/Fold1/train.txt")
test = functools.partial(__reader__, filepath="MQ2007/MQ2007/Fold1/test.txt")
def fetch():
return download(URL, "MQ2007", MD5)
if __name__ == "__main__":
fetch()
mytest = functools.partial(
__reader__, filepath="MQ2007/MQ2007/Fold1/sample", format="listwise")
for label, query in mytest():
print label, query
python/paddle/v2/dataset/sentiment.py 0 → 100644
浏览文件 @ 1f6e0448
# /usr/bin/env python
# -*- coding:utf-8 -*-
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
The script fetch and preprocess movie_reviews data set that provided by NLTK
TODO(yuyang18): Complete dataset.
"""
import collections
from itertools import chain
import nltk
from nltk.corpus import movie_reviews
import paddle.v2.dataset.common
__all__ = ['train', 'test', 'get_word_dict', 'convert']
NUM_TRAINING_INSTANCES = 1600
NUM_TOTAL_INSTANCES = 2000
def download_data_if_not_yet():
"""
Download the data set, if the data set is not download.
"""
try:
# make sure that nltk can find the data
if paddle.v2.dataset.common.DATA_HOME not in nltk.data.path:
nltk.data.path.append(paddle.v2.dataset.common.DATA_HOME)
movie_reviews.categories()
except LookupError:
print "Downloading movie_reviews data set, please wait....."
nltk.download(
'movie_reviews', download_dir=paddle.v2.dataset.common.DATA_HOME)
print "Download data set success....."
print "Path is " + nltk.data.find('corpora/movie_reviews').path
def get_word_dict():
"""
Sorted the words by the frequency of words which occur in sample
:return:
words_freq_sorted
"""
words_freq_sorted = list()
word_freq_dict = collections.defaultdict(int)
download_data_if_not_yet()
for category in movie_reviews.categories():
for field in movie_reviews.fileids(category):
for words in movie_reviews.words(field):
word_freq_dict[words] += 1
words_sort_list = word_freq_dict.items()
words_sort_list.sort(cmp=lambda a, b: b[1] - a[1])
for index, word in enumerate(words_sort_list):
words_freq_sorted.append((word[0], index))
return words_freq_sorted
def sort_files():
"""
Sorted the sample for cross reading the sample
:return:
files_list
"""
files_list = list()
neg_file_list = movie_reviews.fileids('neg')
pos_file_list = movie_reviews.fileids('pos')
files_list = list(chain.from_iterable(zip(neg_file_list, pos_file_list)))
return files_list
def load_sentiment_data():
"""
Load the data set
:return:
data_set
"""
data_set = list()
download_data_if_not_yet()
words_ids = dict(get_word_dict())
for sample_file in sort_files():
words_list = list()
category = 0 if 'neg' in sample_file else 1
for word in movie_reviews.words(sample_file):
words_list.append(words_ids[word.lower()])
data_set.append((words_list, category))
return data_set
def reader_creator(data):
"""
Reader creator, generate an iterator for data set
:param data:
train data set or test data set
"""
for each in data:
yield each[0], each[1]
def train():
"""
Default training set reader creator
"""
data_set = load_sentiment_data()
return reader_creator(data_set[0:NUM_TRAINING_INSTANCES])
def test():
"""
Default test set reader creator
"""
data_set = load_sentiment_data()
return reader_creator(data_set[NUM_TRAINING_INSTANCES:])
def fetch():
nltk.download(
'movie_reviews', download_dir=paddle.v2.dataset.common.DATA_HOME)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, train, 1000, "sentiment_train")
paddle.v2.dataset.common.convert(path, test, 1000, "sentiment_test")
python/paddle/v2/dataset/tests/cifar_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.cifar
import unittest
class TestCIFAR(unittest.TestCase):
def check_reader(self, reader):
sum = 0
label = 0
for l in reader():
self.assertEqual(l[0].size, 3072)
if l[1] > label:
label = l[1]
sum += 1
return sum, label
def test_test10(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.cifar.test10())
self.assertEqual(instances, 10000)
self.assertEqual(max_label_value, 9)
def test_train10(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.cifar.train10())
self.assertEqual(instances, 50000)
self.assertEqual(max_label_value, 9)
def test_test100(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.cifar.test100())
self.assertEqual(instances, 10000)
self.assertEqual(max_label_value, 99)
def test_train100(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.cifar.train100())
self.assertEqual(instances, 50000)
self.assertEqual(max_label_value, 99)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/common_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.common
import unittest
import tempfile
import glob
class TestCommon(unittest.TestCase):
def test_md5file(self):
_, temp_path = tempfile.mkstemp()
with open(temp_path, 'w') as f:
f.write("Hello\n")
self.assertEqual('09f7e02f1290be211da707a266f153b3',
paddle.v2.dataset.common.md5file(temp_path))
def test_download(self):
yi_avatar = 'https://avatars0.githubusercontent.com/u/1548775?v=3&s=460'
self.assertEqual(
paddle.v2.dataset.common.DATA_HOME + '/test/1548775?v=3&s=460',
paddle.v2.dataset.common.download(
yi_avatar, 'test', 'f75287202d6622414c706c36c16f8e0d'))
def test_split(self):
def test_reader():
def reader():
for x in xrange(10):
yield x
return reader
_, temp_path = tempfile.mkstemp()
paddle.v2.dataset.common.split(
test_reader(), 4, suffix=temp_path + '/test-%05d.pickle')
files = glob.glob(temp_path + '/test-%05d.pickle')
self.assertEqual(len(files), 3)
def test_cluster_file_reader(self):
_, temp_path = tempfile.mkstemp()
for x in xrange(5):
with open(temp_path + '/%05d.test' % x) as f:
f.write('%d\n' % x)
reader = paddle.v2.dataset.common.cluster_files_reader(
temp_path + '/*.test', 5, 0)
for idx, e in enumerate(reader()):
self.assertEqual(e, str("0"))
def test_convert(self):
record_num = 10
num_shards = 4
def test_reader():
def reader():
for x in xrange(record_num):
yield x
return reader
path = tempfile.mkdtemp()
paddle.v2.dataset.common.convert(path,
test_reader(), num_shards,
'random_images')
files = glob.glob(path + '/random_images-*')
self.assertEqual(len(files), num_shards)
recs = []
for i in range(0, num_shards):
n = "%s/random_images-%05d-of-%05d" % (path, i, num_shards - 1)
r = recordio.reader(n)
while True:
d = r.read()
if d is None:
break
recs.append(d)
recs.sort()
self.assertEqual(total, record_num)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/flowers_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.flowers
import unittest
class TestFlowers(unittest.TestCase):
def check_reader(self, reader):
sum = 0
label = 0
size = 224 * 224 * 3
for l in reader():
self.assertEqual(l[0].size, size)
if l[1] > label:
label = l[1]
sum += 1
return sum, label
def test_train(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.flowers.train())
self.assertEqual(instances, 6149)
self.assertEqual(max_label_value, 102)
def test_test(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.flowers.test())
self.assertEqual(instances, 1020)
self.assertEqual(max_label_value, 102)
def test_valid(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.flowers.valid())
self.assertEqual(instances, 1020)
self.assertEqual(max_label_value, 102)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/imdb_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.imdb
import unittest
import re
TRAIN_POS_PATTERN = re.compile("aclImdb/train/pos/.*\.txt$")
TRAIN_NEG_PATTERN = re.compile("aclImdb/train/neg/.*\.txt$")
TRAIN_PATTERN = re.compile("aclImdb/train/.*\.txt$")
TEST_POS_PATTERN = re.compile("aclImdb/test/pos/.*\.txt$")
TEST_NEG_PATTERN = re.compile("aclImdb/test/neg/.*\.txt$")
TEST_PATTERN = re.compile("aclImdb/test/.*\.txt$")
class TestIMDB(unittest.TestCase):
word_idx = None
def test_build_dict(self):
if self.word_idx == None:
self.word_idx = paddle.v2.dataset.imdb.build_dict(TRAIN_PATTERN,
150)
self.assertEqual(len(self.word_idx), 7036)
def check_dataset(self, dataset, expected_size):
if self.word_idx == None:
self.word_idx = paddle.v2.dataset.imdb.build_dict(TRAIN_PATTERN,
150)
sum = 0
for l in dataset(self.word_idx):
self.assertEqual(l[1], sum % 2)
sum += 1
self.assertEqual(sum, expected_size)
def test_train(self):
self.check_dataset(paddle.v2.dataset.imdb.train, 25000)
def test_test(self):
self.check_dataset(paddle.v2.dataset.imdb.test, 25000)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/imikolov_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.imikolov
import unittest
WORD_DICT = paddle.v2.dataset.imikolov.build_dict()
class TestMikolov(unittest.TestCase):
def check_reader(self, reader, n):
for l in reader():
self.assertEqual(len(l), n)
def test_train(self):
n = 5
self.check_reader(paddle.v2.dataset.imikolov.train(WORD_DICT, n), n)
first_line = 'aer banknote berlitz calloway centrust cluett fromstein '\
'gitano guterman hydro-quebec ipo kia memotec mlx nahb punts '\
'rake regatta rubens sim snack-food ssangyong swapo wachter'
first_line = [
WORD_DICT.get(ch, WORD_DICT['<unk>'])
for ch in first_line.split(' ')
]
for l in paddle.v2.dataset.imikolov.train(
WORD_DICT, n=-1,
data_type=paddle.v2.dataset.imikolov.DataType.SEQ)():
read_line = l[0][1:]
break
self.assertEqual(first_line, read_line)
def test_test(self):
n = 5
self.check_reader(paddle.v2.dataset.imikolov.test(WORD_DICT, n), n)
first_line = 'consumers may want to move their telephones a little '\
'closer to the tv set'
first_line = [
WORD_DICT.get(ch, WORD_DICT['<unk>'])
for ch in first_line.split(' ')
]
for l in paddle.v2.dataset.imikolov.test(
WORD_DICT, n=-1,
data_type=paddle.v2.dataset.imikolov.DataType.SEQ)():
read_line = l[0][1:]
break
self.assertEqual(first_line, read_line)
def test_total(self):
_, idx = zip(*WORD_DICT.items())
self.assertEqual(sorted(idx)[-1], len(WORD_DICT) - 1)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/mnist_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.mnist
import unittest
class TestMNIST(unittest.TestCase):
def check_reader(self, reader):
sum = 0
label = 0
for l in reader():
self.assertEqual(l[0].size, 784)
if l[1] > label:
label = l[1]
sum += 1
return sum, label
def test_train(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.mnist.train())
self.assertEqual(instances, 60000)
self.assertEqual(max_label_value, 9)
def test_test(self):
instances, max_label_value = self.check_reader(
paddle.v2.dataset.mnist.test())
self.assertEqual(instances, 10000)
self.assertEqual(max_label_value, 9)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/mq2007_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.mq2007
import unittest
class TestMQ2007(unittest.TestCase):
def test_pairwise(self):
for label, query_left, query_right in paddle.v2.dataset.mq2007.test(
format="pairwise"):
self.assertEqual(query_left.shape(), (46, ))
self.assertEqual(query_right.shape(), (46, ))
def test_listwise(self):
for label_array, query_array in paddle.v2.dataset.mq2007.test(
format="listwise"):
self.assertEqual(len(label_array), len(query_array))
if __name__ == "__main__":
unittest.main()
python/paddle/v2/dataset/tests/test_sentiment.py 0 → 100644
浏览文件 @ 1f6e0448
# /usr/bin/env python
# -*- coding:utf-8 -*-
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import nltk
import paddle.v2.dataset.sentiment as st
from nltk.corpus import movie_reviews
class TestSentimentMethods(unittest.TestCase):
def test_get_word_dict(self):
word_dict = st.get_word_dict()[0:10]
test_word_list = [(u',', 0), (u'the', 1), (u'.', 2), (u'a', 3),
(u'and', 4), (u'of', 5), (u'to', 6), (u"'", 7),
(u'is', 8), (u'in', 9)]
for idx, each in enumerate(word_dict):
self.assertEqual(each, test_word_list[idx])
self.assertTrue("/root/.cache/paddle/dataset" in nltk.data.path)
def test_sort_files(self):
last_label = ''
for sample_file in st.sort_files():
current_label = sample_file.split("/")[0]
self.assertNotEqual(current_label, last_label)
last_label = current_label
def test_data_set(self):
data_set = st.load_sentiment_data()
last_label = -1
for each in st.test():
self.assertNotEqual(each[1], last_label)
last_label = each[1]
self.assertEqual(len(data_set), st.NUM_TOTAL_INSTANCES)
self.assertEqual(len(list(st.train())), st.NUM_TRAINING_INSTANCES)
self.assertEqual(
len(list(st.test())),
(st.NUM_TOTAL_INSTANCES - st.NUM_TRAINING_INSTANCES))
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/voc2012_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.voc2012
import unittest
class TestVOC(unittest.TestCase):
def check_reader(self, reader):
sum = 0
label = 0
for l in reader():
self.assertEqual(l[0].size, 3 * l[1].size)
sum += 1
return sum
def test_train(self):
count = self.check_reader(paddle.v2.dataset.voc_seg.train())
self.assertEqual(count, 2913)
def test_test(self):
count = self.check_reader(paddle.v2.dataset.voc_seg.test())
self.assertEqual(count, 1464)
def test_val(self):
count = self.check_reader(paddle.v2.dataset.voc_seg.val())
self.assertEqual(count, 1449)
if __name__ == '__main__':
unittest.main()
python/paddle/v2/dataset/tests/wmt16_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import paddle.v2.dataset.wmt16
import unittest
class TestWMT16(unittest.TestCase):
def checkout_one_sample(self, sample):
# train data has 3 field: source language word indices,
# target language word indices, and target next word indices.
self.assertEqual(len(sample), 3)
# test start mark and end mark in source word indices.
self.assertEqual(sample[0][0], 0)
self.assertEqual(sample[0][-1], 1)
# test start mask in target word indices
self.assertEqual(sample[1][0], 0)
# test en mask in target next word indices
self.assertEqual(sample[2][-1], 1)
def test_train(self):
for idx, sample in enumerate(
paddle.v2.dataset.wmt16.train(
src_dict_size=100000, trg_dict_size=100000)()):
if idx >= 10: break
self.checkout_one_sample(sample)
def test_test(self):
for idx, sample in enumerate(
paddle.v2.dataset.wmt16.test(
src_dict_size=1000, trg_dict_size=1000)()):
if idx >= 10: break
self.checkout_one_sample(sample)
def test_val(self):
for idx, sample in enumerate(
paddle.v2.dataset.wmt16.validation(
src_dict_size=1000, trg_dict_size=1000)()):
if idx >= 10: break
self.checkout_one_sample(sample)
def test_get_dict(self):
dict_size = 1000
word_dict = paddle.v2.dataset.wmt16.get_dict("en", dict_size, True)
self.assertEqual(len(word_dict), dict_size)
self.assertEqual(word_dict[0], "<s>")
self.assertEqual(word_dict[1], "<e>")
self.assertEqual(word_dict[2], "<unk>")
if __name__ == "__main__":
unittest.main()
python/paddle/v2/dataset/uci_housing.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
UCI Housing dataset.
This module will download dataset from
https://archive.ics.uci.edu/ml/machine-learning-databases/housing/ and
parse training set and test set into paddle reader creators.
"""
import numpy as np
import os
import paddle.v2.dataset.common
from paddle.v2.parameters import Parameters
__all__ = ['train', 'test']
URL = 'https://archive.ics.uci.edu/ml/machine-learning-databases/housing/housing.data'
MD5 = 'd4accdce7a25600298819f8e28e8d593'
feature_names = [
'CRIM', 'ZN', 'INDUS', 'CHAS', 'NOX', 'RM', 'AGE', 'DIS', 'RAD', 'TAX',
'PTRATIO', 'B', 'LSTAT', 'convert'
]
UCI_TRAIN_DATA = None
UCI_TEST_DATA = None
URL_MODEL = 'https://github.com/PaddlePaddle/book/raw/develop/01.fit_a_line/fit_a_line.tar'
MD5_MODEL = '52fc3da8ef3937822fcdd87ee05c0c9b'
def feature_range(maximums, minimums):
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
fig, ax = plt.subplots()
feature_num = len(maximums)
ax.bar(range(feature_num), maximums - minimums, color='r', align='center')
ax.set_title('feature scale')
plt.xticks(range(feature_num), feature_names)
plt.xlim([-1, feature_num])
fig.set_figheight(6)
fig.set_figwidth(10)
if not os.path.exists('./image'):
os.makedirs('./image')
fig.savefig('image/ranges.png', dpi=48)
plt.close(fig)
def load_data(filename, feature_num=14, ratio=0.8):
global UCI_TRAIN_DATA, UCI_TEST_DATA
if UCI_TRAIN_DATA is not None and UCI_TEST_DATA is not None:
return
data = np.fromfile(filename, sep=' ')
data = data.reshape(data.shape[0] / feature_num, feature_num)
maximums, minimums, avgs = data.max(axis=0), data.min(axis=0), data.sum(
axis=0) / data.shape[0]
feature_range(maximums[:-1], minimums[:-1])
for i in xrange(feature_num - 1):
data[:, i] = (data[:, i] - avgs[i]) / (maximums[i] - minimums[i])
offset = int(data.shape[0] * ratio)
UCI_TRAIN_DATA = data[:offset]
UCI_TEST_DATA = data[offset:]
def train():
"""
UCI_HOUSING training set creator.
It returns a reader creator, each sample in the reader is features after
normalization and price number.
:return: Training reader creator
:rtype: callable
"""
global UCI_TRAIN_DATA
load_data(paddle.v2.dataset.common.download(URL, 'uci_housing', MD5))
def reader():
for d in UCI_TRAIN_DATA:
yield d[:-1], d[-1:]
return reader
def test():
"""
UCI_HOUSING test set creator.
It returns a reader creator, each sample in the reader is features after
normalization and price number.
:return: Test reader creator
:rtype: callable
"""
global UCI_TEST_DATA
load_data(paddle.v2.dataset.common.download(URL, 'uci_housing', MD5))
def reader():
for d in UCI_TEST_DATA:
yield d[:-1], d[-1:]
return reader
def model():
tar_file = paddle.v2.dataset.common.download(URL_MODEL, 'fit_a_line.tar',
MD5_MODEL)
with open(tar_file, 'r') as f:
parameters = Parameters.from_tar(f)
return parameters
def fetch():
paddle.v2.dataset.common.download(URL, 'uci_housing', MD5)
def convert(path):
"""
Converts dataset to recordio format
"""
paddle.v2.dataset.common.convert(path, train(), 1000, "uci_housing_train")
paddle.v2.dataset.common.convert(path, test(), 1000, "uci_houseing_test")
python/paddle/v2/dataset/voc2012.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Image dataset for segmentation.
The 2012 dataset contains images from 2008-2011 for which additional
segmentations have been prepared. As in previous years the assignment
to training/test sets has been maintained. The total number of images
with segmentation has been increased from 7,062 to 9,993.
"""
import tarfile
import io
import numpy as np
from paddle.v2.dataset.common import download
from paddle.v2.image import *
from PIL import Image
__all__ = ['train', 'test', 'val']
VOC_URL = 'http://host.robots.ox.ac.uk/pascal/VOC/voc2012/\
VOCtrainval_11-May-2012.tar'
VOC_MD5 = '6cd6e144f989b92b3379bac3b3de84fd'
SET_FILE = 'VOCdevkit/VOC2012/ImageSets/Segmentation/{}.txt'
DATA_FILE = 'VOCdevkit/VOC2012/JPEGImages/{}.jpg'
LABEL_FILE = 'VOCdevkit/VOC2012/SegmentationClass/{}.png'
CACHE_DIR = 'voc2012'
def reader_creator(filename, sub_name):
tarobject = tarfile.open(filename)
name2mem = {}
for ele in tarobject.getmembers():
name2mem[ele.name] = ele
def reader():
set_file = SET_FILE.format(sub_name)
sets = tarobject.extractfile(name2mem[set_file])
for line in sets:
line = line.strip()
data_file = DATA_FILE.format(line)
label_file = LABEL_FILE.format(line)
data = tarobject.extractfile(name2mem[data_file]).read()
label = tarobject.extractfile(name2mem[label_file]).read()
data = Image.open(io.BytesIO(data))
label = Image.open(io.BytesIO(label))
data = np.array(data)
label = np.array(label)
yield data, label
return reader
def train():
"""
Create a train dataset reader containing 2913 images in HWC order.
"""
return reader_creator(download(VOC_URL, CACHE_DIR, VOC_MD5), 'trainval')
def test():
"""
Create a test dataset reader containing 1464 images in HWC order.
"""
return reader_creator(download(VOC_URL, CACHE_DIR, VOC_MD5), 'train')
def val():
"""
Create a val dataset reader containing 1449 images in HWC order.
"""
return reader_creator(download(VOC_URL, CACHE_DIR, VOC_MD5), 'val')
python/paddle/v2/dataset/wmt14.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
WMT14 dataset.
The original WMT14 dataset is too large and a small set of data for set is
provided. This module will download dataset from
http://paddlepaddle.cdn.bcebos.com/demo/wmt_shrinked_data/wmt14.tgz and
parse training set and test set into paddle reader creators.
"""
import tarfile
import gzip
import paddle.v2.dataset.common
from paddle.v2.parameters import Parameters
__all__ = [
'train',
'test',
'get_dict',
'convert',
]
URL_DEV_TEST = ('http://www-lium.univ-lemans.fr/~schwenk/'
'cslm_joint_paper/data/dev+test.tgz')
MD5_DEV_TEST = '7d7897317ddd8ba0ae5c5fa7248d3ff5'
# this is a small set of data for test. The original data is too large and
# will be add later.
URL_TRAIN = ('http://paddlepaddle.cdn.bcebos.com/demo/'
'wmt_shrinked_data/wmt14.tgz')
MD5_TRAIN = '0791583d57d5beb693b9414c5b36798c'
# BLEU of this trained model is 26.92
URL_MODEL = 'http://paddlepaddle.bj.bcebos.com/demo/wmt_14/wmt14_model.tar.gz'
MD5_MODEL = '0cb4a5366189b6acba876491c8724fa3'
START = "<s>"
END = "<e>"
UNK = "<unk>"
UNK_IDX = 2
def __read_to_dict(tar_file, dict_size):
def __to_dict(fd, size):
out_dict = dict()
for line_count, line in enumerate(fd):
if line_count < size:
out_dict[line.strip()] = line_count
else:
break
return out_dict
with tarfile.open(tar_file, mode='r') as f:
names = [
each_item.name for each_item in f
if each_item.name.endswith("src.dict")
]
assert len(names) == 1
src_dict = __to_dict(f.extractfile(names[0]), dict_size)
names = [
each_item.name for each_item in f
if each_item.name.endswith("trg.dict")
]
assert len(names) == 1
trg_dict = __to_dict(f.extractfile(names[0]), dict_size)
return src_dict, trg_dict
def reader_creator(tar_file, file_name, dict_size):
def reader():
src_dict, trg_dict = __read_to_dict(tar_file, dict_size)
with tarfile.open(tar_file, mode='r') as f:
names = [
each_item.name for each_item in f
if each_item.name.endswith(file_name)
]
for name in names:
for line in f.extractfile(name):
line_split = line.strip().split('\t')
if len(line_split) != 2:
continue
src_seq = line_split[0] # one source sequence
src_words = src_seq.split()
src_ids = [
src_dict.get(w, UNK_IDX)
for w in [START] + src_words + [END]
]
trg_seq = line_split[1] # one target sequence
trg_words = trg_seq.split()
trg_ids = [trg_dict.get(w, UNK_IDX) for w in trg_words]
# remove sequence whose length > 80 in training mode
if len(src_ids) > 80 or len(trg_ids) > 80:
continue
trg_ids_next = trg_ids + [trg_dict[END]]
trg_ids = [trg_dict[START]] + trg_ids
yield src_ids, trg_ids, trg_ids_next
return reader
def train(dict_size):
"""
WMT14 training set creator.
It returns a reader creator, each sample in the reader is source language
word ID sequence, target language word ID sequence and next word ID
sequence.
:return: Training reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN),
'train/train', dict_size)
def test(dict_size):
"""
WMT14 test set creator.
It returns a reader creator, each sample in the reader is source language
word ID sequence, target language word ID sequence and next word ID
sequence.
:return: Test reader creator
:rtype: callable
"""
return reader_creator(
paddle.v2.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN),
'test/test', dict_size)
def gen(dict_size):
return reader_creator(
paddle.v2.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN),
'gen/gen', dict_size)
def model():
tar_file = paddle.v2.dataset.common.download(URL_MODEL, 'wmt14', MD5_MODEL)
with gzip.open(tar_file, 'r') as f:
parameters = Parameters.from_tar(f)
return parameters
def get_dict(dict_size, reverse=True):
# if reverse = False, return dict = {'a':'001', 'b':'002', ...}
# else reverse = true, return dict = {'001':'a', '002':'b', ...}
tar_file = paddle.v2.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN)
src_dict, trg_dict = __read_to_dict(tar_file, dict_size)
if reverse:
src_dict = {v: k for k, v in src_dict.items()}
trg_dict = {v: k for k, v in trg_dict.items()}
return src_dict, trg_dict
def fetch():
paddle.v2.dataset.common.download(URL_TRAIN, 'wmt14', MD5_TRAIN)
paddle.v2.dataset.common.download(URL_MODEL, 'wmt14', MD5_MODEL)
def convert(path):
"""
Converts dataset to recordio format
"""
dict_size = 30000
paddle.v2.dataset.common.convert(path,
train(dict_size), 1000, "wmt14_train")
paddle.v2.dataset.common.convert(path, test(dict_size), 1000, "wmt14_test")
python/paddle/v2/dataset/wmt16.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
ACL2016 Multimodal Machine Translation. Please see this website for more
details: http://www.statmt.org/wmt16/multimodal-task.html#task1
If you use the dataset created for your task, please cite the following paper:
Multi30K: Multilingual English-German Image Descriptions.
@article{elliott-EtAl:2016:VL16,
author = {{Elliott}, D. and {Frank}, S. and {Sima"an}, K. and {Specia}, L.},
title = {Multi30K: Multilingual English-German Image Descriptions},
booktitle = {Proceedings of the 6th Workshop on Vision and Language},
year = {2016},
pages = {70--74},
year = 2016
}
"""
import os
import tarfile
import gzip
from collections import defaultdict
import paddle.v2.dataset.common
__all__ = [
"train",
"test",
"validation",
"convert",
"fetch",
"get_dict",
]
DATA_URL = ("http://cloud.dlnel.org/filepub/"
"?uuid=46a0808e-ddd8-427c-bacd-0dbc6d045fed")
DATA_MD5 = "0c38be43600334966403524a40dcd81e"
TOTAL_EN_WORDS = 11250
TOTAL_DE_WORDS = 19220
START_MARK = "<s>"
END_MARK = "<e>"
UNK_MARK = "<unk>"
def __build_dict(tar_file, dict_size, save_path, lang):
word_dict = defaultdict(int)
with tarfile.open(tar_file, mode="r") as f:
for line in f.extractfile("wmt16/train"):
line_split = line.strip().split("\t")
if len(line_split) != 2: continue
sen = line_split[0] if lang == "en" else line_split[1]
for w in sen.split():
word_dict[w] += 1
with open(save_path, "w") as fout:
fout.write("%s\n%s\n%s\n" % (START_MARK, END_MARK, UNK_MARK))
for idx, word in enumerate(
sorted(
word_dict.iteritems(), key=lambda x: x[1], reverse=True)):
if idx + 3 == dict_size: break
fout.write("%s\n" % (word[0]))
def __load_dict(tar_file, dict_size, lang, reverse=False):
dict_path = os.path.join(paddle.v2.dataset.common.DATA_HOME,
"wmt16/%s_%d.dict" % (lang, dict_size))
if not os.path.exists(dict_path) or (
len(open(dict_path, "r").readlines()) != dict_size):
__build_dict(tar_file, dict_size, dict_path, lang)
word_dict = {}
with open(dict_path, "r") as fdict:
for idx, line in enumerate(fdict):
if reverse:
word_dict[idx] = line.strip()
else:
word_dict[line.strip()] = idx
return word_dict
def __get_dict_size(src_dict_size, trg_dict_size, src_lang):
src_dict_size = min(src_dict_size, (TOTAL_EN_WORDS if src_lang == "en" else
TOTAL_DE_WORDS))
trg_dict_size = min(trg_dict_size, (TOTAL_DE_WORDS if src_lang == "en" else
TOTAL_ENG_WORDS))
return src_dict_size, trg_dict_size
def reader_creator(tar_file, file_name, src_dict_size, trg_dict_size, src_lang):
def reader():
src_dict = __load_dict(tar_file, src_dict_size, src_lang)
trg_dict = __load_dict(tar_file, trg_dict_size,
("de" if src_lang == "en" else "en"))
# the indice for start mark, end mark, and unk are the same in source
# language and target language. Here uses the source language
# dictionary to determine their indices.
start_id = src_dict[START_MARK]
end_id = src_dict[END_MARK]
unk_id = src_dict[UNK_MARK]
src_col = 0 if src_lang == "en" else 1
trg_col = 1 - src_col
with tarfile.open(tar_file, mode="r") as f:
for line in f.extractfile(file_name):
line_split = line.strip().split("\t")
if len(line_split) != 2:
continue
src_words = line_split[src_col].split()
src_ids = [start_id] + [
src_dict.get(w, unk_id) for w in src_words
] + [end_id]
trg_words = line_split[trg_col].split()
trg_ids = [trg_dict.get(w, unk_id) for w in trg_words]
trg_ids_next = trg_ids + [end_id]
trg_ids = [start_id] + trg_ids
yield src_ids, trg_ids, trg_ids_next
return reader
def train(src_dict_size, trg_dict_size, src_lang="en"):
"""
WMT16 train set reader.
This function returns the reader for train data. Each sample the reader
returns is made up of three fields: the source language word index sequence,
target language word index sequence and next word index sequence.
NOTE:
The original like for training data is:
http://www.quest.dcs.shef.ac.uk/wmt16_files_mmt/training.tar.gz
paddle.dataset.wmt16 provides a tokenized version of the original dataset by
using moses's tokenization script:
https://github.com/moses-smt/mosesdecoder/blob/master/scripts/tokenizer/tokenizer.perl
Args:
src_dict_size(int): Size of the source language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
trg_dict_size(int): Size of the target language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
src_lang(string): A string indicating which language is the source
language. Available options are: "en" for English
and "de" for Germany.
Returns:
callable: The train reader.
"""
if src_lang not in ["en", "de"]:
raise ValueError("An error language type. Only support: "
"en (for English); de(for Germany).")
src_dict_size, trg_dict_size = __get_dict_size(src_dict_size, trg_dict_size,
src_lang)
return reader_creator(
tar_file=paddle.v2.dataset.common.download(DATA_URL, "wmt16", DATA_MD5,
"wmt16.tar.gz"),
file_name="wmt16/train",
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang)
def test(src_dict_size, trg_dict_size, src_lang="en"):
"""
WMT16 test set reader.
This function returns the reader for test data. Each sample the reader
returns is made up of three fields: the source language word index sequence,
target language word index sequence and next word index sequence.
NOTE:
The original like for test data is:
http://www.quest.dcs.shef.ac.uk/wmt16_files_mmt/mmt16_task1_test.tar.gz
paddle.dataset.wmt16 provides a tokenized version of the original dataset by
using moses's tokenization script:
https://github.com/moses-smt/mosesdecoder/blob/master/scripts/tokenizer/tokenizer.perl
Args:
src_dict_size(int): Size of the source language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
trg_dict_size(int): Size of the target language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
src_lang(string): A string indicating which language is the source
language. Available options are: "en" for English
and "de" for Germany.
Returns:
callable: The test reader.
"""
if src_lang not in ["en", "de"]:
raise ValueError("An error language type. "
"Only support: en (for English); de(for Germany).")
src_dict_size, trg_dict_size = __get_dict_size(src_dict_size, trg_dict_size,
src_lang)
return reader_creator(
tar_file=paddle.v2.dataset.common.download(DATA_URL, "wmt16", DATA_MD5,
"wmt16.tar.gz"),
file_name="wmt16/test",
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang)
def validation(src_dict_size, trg_dict_size, src_lang="en"):
"""
WMT16 validation set reader.
This function returns the reader for validation data. Each sample the reader
returns is made up of three fields: the source language word index sequence,
target language word index sequence and next word index sequence.
NOTE:
The original like for validation data is:
http://www.quest.dcs.shef.ac.uk/wmt16_files_mmt/validation.tar.gz
paddle.dataset.wmt16 provides a tokenized version of the original dataset by
using moses's tokenization script:
https://github.com/moses-smt/mosesdecoder/blob/master/scripts/tokenizer/tokenizer.perl
Args:
src_dict_size(int): Size of the source language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
trg_dict_size(int): Size of the target language dictionary. Three
special tokens will be added into the dictionary:
<s> for start mark, <e> for end mark, and <unk> for
unknown word.
src_lang(string): A string indicating which language is the source
language. Available options are: "en" for English
and "de" for Germany.
Returns:
callable: The validation reader.
"""
if src_lang not in ["en", "de"]:
raise ValueError("An error language type. "
"Only support: en (for English); de(for Germany).")
src_dict_size, trg_dict_size = __get_dict_size(src_dict_size, trg_dict_size,
src_lang)
return reader_creator(
tar_file=paddle.v2.dataset.common.download(DATA_URL, "wmt16", DATA_MD5,
"wmt16.tar.gz"),
file_name="wmt16/val",
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang)
def get_dict(lang, dict_size, reverse=False):
"""
return the word dictionary for the specified language.
Args:
lang(string): A string indicating which language is the source
language. Available options are: "en" for English
and "de" for Germany.
dict_size(int): Size of the specified language dictionary.
reverse(bool): If reverse is set to False, the returned python
dictionary will use word as key and use index as value.
If reverse is set to True, the returned python
dictionary will use index as key and word as value.
Returns:
dict: The word dictionary for the specific language.
"""
if lang == "en": dict_size = min(dict_size, TOTAL_EN_WORDS)
else: dict_size = min(dict_size, TOTAL_DE_WORDS)
dict_path = os.path.join(paddle.v2.dataset.common.DATA_HOME,
"wmt16/%s_%d.dict" % (lang, dict_size))
assert os.path.exists(dict_path), "Word dictionary does not exist. "
"Please invoke paddle.dataset.wmt16.train/test/validation first "
"to build the dictionary."
tar_file = os.path.join(paddle.v2.dataset.common.DATA_HOME, "wmt16.tar.gz")
return __load_dict(tar_file, dict_size, lang, reverse)
def fetch():
"""download the entire dataset.
"""
paddle.v4.dataset.common.download(DATA_URL, "wmt16", DATA_MD5,
"wmt16.tar.gz")
def convert(path, src_dict_size, trg_dict_size, src_lang):
"""Converts dataset to recordio format.
"""
paddle.v2.dataset.common.convert(
path,
train(
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang),
1000,
"wmt16_train")
paddle.v2.dataset.common.convert(
path,
test(
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang),
1000,
"wmt16_test")
paddle.v2.dataset.common.convert(
path,
validation(
src_dict_size=src_dict_size,
trg_dict_size=trg_dict_size,
src_lang=src_lang),
1000,
"wmt16_validation")
python/paddle/v2/image.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
This file contains some common interfaces for image preprocess.
Many users are confused about the image layout. We introduce
the image layout as follows.
- CHW Layout
- The abbreviations: C=channel, H=Height, W=Width
- The default layout of image opened by cv2 or PIL is HWC.
PaddlePaddle only supports the CHW layout. And CHW is simply
a transpose of HWC. It must transpose the input image.
- Color format: RGB or BGR
OpenCV use BGR color format. PIL use RGB color format. Both
formats can be used for training. Noted that, the format should
be keep consistent between the training and inference peroid.
"""
import numpy as np
try:
import cv2
except ImportError:
cv2 = None
import os
import tarfile
import cPickle
__all__ = [
"load_image_bytes", "load_image", "resize_short", "to_chw", "center_crop",
"random_crop", "left_right_flip", "simple_transform", "load_and_transform",
"batch_images_from_tar"
]
def batch_images_from_tar(data_file,
dataset_name,
img2label,
num_per_batch=1024):
"""
Read images from tar file and batch them into batch file.
:param data_file: path of image tar file
:type data_file: string
:param dataset_name: 'train','test' or 'valid'
:type dataset_name: string
:param img2label: a dic with image file name as key
and image's label as value
:type img2label: dic
:param num_per_batch: image number per batch file
:type num_per_batch: int
:return: path of list file containing paths of batch file
:rtype: string
"""
batch_dir = data_file + "_batch"
out_path = "%s/%s" % (batch_dir, dataset_name)
meta_file = "%s/%s.txt" % (batch_dir, dataset_name)
if os.path.exists(out_path):
return meta_file
else:
os.makedirs(out_path)
tf = tarfile.open(data_file)
mems = tf.getmembers()
data = []
labels = []
file_id = 0
for mem in mems:
if mem.name in img2label:
data.append(tf.extractfile(mem).read())
labels.append(img2label[mem.name])
if len(data) == num_per_batch:
output = {}
output['label'] = labels
output['data'] = data
cPickle.dump(
output,
open('%s/batch_%d' % (out_path, file_id), 'w'),
protocol=cPickle.HIGHEST_PROTOCOL)
file_id += 1
data = []
labels = []
if len(data) > 0:
output = {}
output['label'] = labels
output['data'] = data
cPickle.dump(
output,
open('%s/batch_%d' % (out_path, file_id), 'w'),
protocol=cPickle.HIGHEST_PROTOCOL)
with open(meta_file, 'a') as meta:
for file in os.listdir(out_path):
meta.write(os.path.abspath("%s/%s" % (out_path, file)) + "\n")
return meta_file
def load_image_bytes(bytes, is_color=True):
"""
Load an color or gray image from bytes array.
Example usage:
.. code-block:: python
with open('cat.jpg') as f:
im = load_image_bytes(f.read())
:param bytes: the input image bytes array.
:type bytes: str
:param is_color: If set is_color True, it will load and
return a color image. Otherwise, it will
load and return a gray image.
:type is_color: bool
"""
flag = 1 if is_color else 0
file_bytes = np.asarray(bytearray(bytes), dtype=np.uint8)
img = cv2.imdecode(file_bytes, flag)
return img
def load_image(file, is_color=True):
"""
Load an color or gray image from the file path.
Example usage:
.. code-block:: python
im = load_image('cat.jpg')
:param file: the input image path.
:type file: string
:param is_color: If set is_color True, it will load and
return a color image. Otherwise, it will
load and return a gray image.
:type is_color: bool
"""
# cv2.IMAGE_COLOR for OpenCV3
# cv2.CV_LOAD_IMAGE_COLOR for older OpenCV Version
# cv2.IMAGE_GRAYSCALE for OpenCV3
# cv2.CV_LOAD_IMAGE_GRAYSCALE for older OpenCV Version
# Here, use constant 1 and 0
# 1: COLOR, 0: GRAYSCALE
flag = 1 if is_color else 0
im = cv2.imread(file, flag)
return im
def resize_short(im, size):
"""
Resize an image so that the length of shorter edge is size.
Example usage:
.. code-block:: python
im = load_image('cat.jpg')
im = resize_short(im, 256)
:param im: the input image with HWC layout.
:type im: ndarray
:param size: the shorter edge size of image after resizing.
:type size: int
"""
h, w = im.shape[:2]
h_new, w_new = size, size
if h > w:
h_new = size * h / w
else:
w_new = size * w / h
im = cv2.resize(im, (h_new, w_new), interpolation=cv2.INTER_CUBIC)
return im
def to_chw(im, order=(2, 0, 1)):
"""
Transpose the input image order. The image layout is HWC format
opened by cv2 or PIL. Transpose the input image to CHW layout
according the order (2,0,1).
Example usage:
.. code-block:: python
im = load_image('cat.jpg')
im = resize_short(im, 256)
im = to_chw(im)
:param im: the input image with HWC layout.
:type im: ndarray
:param order: the transposed order.
:type order: tuple|list
"""
assert len(im.shape) == len(order)
im = im.transpose(order)
return im
def center_crop(im, size, is_color=True):
"""
Crop the center of image with size.
Example usage:
.. code-block:: python
im = center_crop(im, 224)
:param im: the input image with HWC layout.
:type im: ndarray
:param size: the cropping size.
:type size: int
:param is_color: whether the image is color or not.
:type is_color: bool
"""
h, w = im.shape[:2]
h_start = (h - size) / 2
w_start = (w - size) / 2
h_end, w_end = h_start + size, w_start + size
if is_color:
im = im[h_start:h_end, w_start:w_end, :]
else:
im = im[h_start:h_end, w_start:w_end]
return im
def random_crop(im, size, is_color=True):
"""
Randomly crop input image with size.
Example usage:
.. code-block:: python
im = random_crop(im, 224)
:param im: the input image with HWC layout.
:type im: ndarray
:param size: the cropping size.
:type size: int
:param is_color: whether the image is color or not.
:type is_color: bool
"""
h, w = im.shape[:2]
h_start = np.random.randint(0, h - size + 1)
w_start = np.random.randint(0, w - size + 1)
h_end, w_end = h_start + size, w_start + size
if is_color:
im = im[h_start:h_end, w_start:w_end, :]
else:
im = im[h_start:h_end, w_start:w_end]
return im
def left_right_flip(im, is_color=True):
"""
Flip an image along the horizontal direction.
Return the flipped image.
Example usage:
.. code-block:: python
im = left_right_flip(im)
:param im: input image with HWC layout or HW layout for gray image
:type im: ndarray
:param is_color: whether input image is color or not
:type is_color: bool
"""
if len(im.shape) == 3 and is_color:
return im[:, ::-1, :]
else:
return im[:, ::-1]
def simple_transform(im,
resize_size,
crop_size,
is_train,
is_color=True,
mean=None):
"""
Simply data argumentation for training. These operations include
resizing, croping and flipping.
Example usage:
.. code-block:: python
im = simple_transform(im, 256, 224, True)
:param im: The input image with HWC layout.
:type im: ndarray
:param resize_size: The shorter edge length of the resized image.
:type resize_size: int
:param crop_size: The cropping size.
:type crop_size: int
:param is_train: Whether it is training or not.
:type is_train: bool
:param is_color: whether the image is color or not.
:type is_color: bool
:param mean: the mean values, which can be element-wise mean values or
mean values per channel.
:type mean: numpy array | list
"""
im = resize_short(im, resize_size)
if is_train:
im = random_crop(im, crop_size, is_color=is_color)
if np.random.randint(2) == 0:
im = left_right_flip(im, is_color)
else:
im = center_crop(im, crop_size, is_color)
im = center_crop(im, crop_size, is_color=is_color)
if len(im.shape) == 3:
im = to_chw(im)
im = im.astype('float32')
if mean is not None:
mean = np.array(mean, dtype=np.float32)
# mean value, may be one value per channel
if mean.ndim == 1 and is_color:
mean = mean[:, np.newaxis, np.newaxis]
elif mean.ndim == 1:
mean = mean
else:
# elementwise mean
assert len(mean.shape) == len(im)
im -= mean
return im
def load_and_transform(filename,
resize_size,
crop_size,
is_train,
is_color=True,
mean=None):
"""
Load image from the input file `filename` and transform image for
data argumentation. Please refer to the `simple_transform` interface
for the transform operations.
Example usage:
.. code-block:: python
im = load_and_transform('cat.jpg', 256, 224, True)
:param filename: The file name of input image.
:type filename: string
:param resize_size: The shorter edge length of the resized image.
:type resize_size: int
:param crop_size: The cropping size.
:type crop_size: int
:param is_train: Whether it is training or not.
:type is_train: bool
:param is_color: whether the image is color or not.
:type is_color: bool
:param mean: the mean values, which can be element-wise mean values or
mean values per channel.
:type mean: numpy array | list
"""
im = load_image(filename, is_color)
im = simple_transform(im, resize_size, crop_size, is_train, is_color, mean)
return im
python/paddle/v2/minibatch.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
__all__ = ['batch']
def batch(reader, batch_size):
"""
Create a batched reader.
:param reader: the data reader to read from.
:type reader: callable
:param batch_size: size of each mini-batch
:type batch_size: int
:return: the batched reader.
:rtype: callable
"""
def batch_reader():
r = reader()
b = []
for instance in r:
b.append(instance)
if len(b) == batch_size:
yield b
b = []
if b:
yield b
return batch_reader
python/paddle/v2/reader/__init__.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
At training and testing time, PaddlePaddle programs need to read data. To ease
the users' work to write data reading code, we define that
- A *reader* is a function that reads data (from file, network, random number
generator, etc) and yields data items.
- A *reader creator* is a function that returns a reader function.
- A *reader decorator* is a function, which accepts one or more readers, and
returns a reader.
- A *batch reader* is a function that reads data (from *reader*, file, network,
random number generator, etc) and yields a batch of data items.
#####################
Data Reader Interface
#####################
Indeed, *data reader* doesn't have to be a function that reads and yields data
items. It can be any function with no parameter that creates a iterable
(anything can be used in :code:`for x in iterable`)\:
.. code-block:: python
iterable = data_reader()
Element produced from the iterable should be a **single** entry of data,
**not** a mini batch. That entry of data could be a single item, or a tuple of
items.
Item should be of `supported type <http://www.paddlepaddle.org/doc/ui/data_provider
/pydataprovider2.html?highlight=dense_vector#input-types>`_ (e.g., numpy 1d
array of float32, int, list of int)
An example implementation for single item data reader creator:
.. code-block:: python
def reader_creator_random_image(width, height):
def reader():
while True:
yield numpy.random.uniform(-1, 1, size=width*height)
return reader
An example implementation for multiple item data reader creator:
.. code-block:: python
def reader_creator_random_image_and_label(width, height, label):
def reader():
while True:
yield numpy.random.uniform(-1, 1, size=width*height), label
return reader
TODO(yuyang18): Should we add whole design doc here?
"""
import decorator
from decorator import *
import creator
__all__ = decorator.__all__ + ['creator']
python/paddle/v2/reader/creator.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
"""
Creator package contains some simple reader creator, which could
be used in user program.
"""
__all__ = ['np_array', 'text_file', 'recordio', 'cloud_reader']
def np_array(x):
"""
Creates a reader that yields elements of x, if it is a
numpy vector. Or rows of x, if it is a numpy matrix.
Or any sub-hyperplane indexed by the highest dimension.
:param x: the numpy array to create reader from.
:returns: data reader created from x.
"""
def reader():
if x.ndim < 1:
yield x
for e in x:
yield e
return reader
def text_file(path):
"""
Creates a data reader that outputs text line by line from given text file.
Trailing new line ('\\\\n') of each line will be removed.
:path: path of the text file.
:returns: data reader of text file
"""
def reader():
f = open(path, "r")
for l in f:
yield l.rstrip('\n')
f.close()
return reader
def recordio(paths, buf_size=100):
"""
Creates a data reader from given RecordIO file paths separated by ",",
glob pattern is supported.
:path: path of recordio files, can be a string or a string list.
:returns: data reader of recordio files.
"""
import recordio as rec
import paddle.v2.reader.decorator as dec
import cPickle as pickle
def reader():
if isinstance(paths, basestring):
path = paths
else:
path = ",".join(paths)
f = rec.reader(path)
while True:
r = f.read()
if r is None:
break
yield pickle.loads(r)
f.close()
return dec.buffered(reader, buf_size)
pass_num = 0
def cloud_reader(paths, etcd_endpoints, timeout_sec=5, buf_size=64):
"""
Create a data reader that yield a record one by one from
the paths:
:paths: path of recordio files, can be a string or a string list.
:etcd_endpoints: the endpoints for etcd cluster
:returns: data reader of recordio files.
.. code-block:: python
from paddle.v2.reader.creator import cloud_reader
etcd_endpoints = "http://127.0.0.1:2379"
trainer.train.(
reader=cloud_reader(["/work/dataset/uci_housing/uci_housing*"], etcd_endpoints),
)
"""
import os
import cPickle as pickle
import paddle.v2.master as master
c = master.client(etcd_endpoints, timeout_sec, buf_size)
if isinstance(paths, basestring):
path = [paths]
else:
path = paths
c.set_dataset(path)
def reader():
global pass_num
c.paddle_start_get_records(pass_num)
pass_num += 1
while True:
r, e = c.next_record()
if not r:
if e != -2:
print "get record error: ", e
break
yield pickle.loads(r)
return reader
python/paddle/v2/reader/decorator.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
__all__ = [
'map_readers', 'buffered', 'compose', 'chain', 'shuffle',
'ComposeNotAligned', 'firstn', 'xmap_readers', 'PipeReader'
]
from threading import Thread
import subprocess
from Queue import Queue
import itertools
import random
import zlib
def map_readers(func, *readers):
"""
Creates a data reader that outputs return value of function using
output of each data readers as arguments.
:param func: function to use. The type of func should be (Sample) => Sample
:type: callable
:param readers: readers whose outputs will be used as arguments of func.
:return: the created data reader.
:rtype: callable
"""
def reader():
rs = []
for r in readers:
rs.append(r())
for e in itertools.imap(func, *rs):
yield e
return reader
def shuffle(reader, buf_size):
"""
Creates a data reader whose data output is shuffled.
Output from the iterator that created by original reader will be
buffered into shuffle buffer, and then shuffled. The size of shuffle buffer
is determined by argument buf_size.
:param reader: the original reader whose output will be shuffled.
:type reader: callable
:param buf_size: shuffle buffer size.
:type buf_size: int
:return: the new reader whose output is shuffled.
:rtype: callable
"""
def data_reader():
buf = []
for e in reader():
buf.append(e)
if len(buf) >= buf_size:
random.shuffle(buf)
for b in buf:
yield b
buf = []
if len(buf) > 0:
random.shuffle(buf)
for b in buf:
yield b
return data_reader
def chain(*readers):
"""
Creates a data reader whose output is the outputs of input data
readers chained together.
If input readers output following data entries:
[0, 0, 0]
[1, 1, 1]
[2, 2, 2]
The chained reader will output:
[0, 0, 0, 1, 1, 1, 2, 2, 2]
:param readers: input readers.
:return: the new data reader.
:rtype: callable
"""
def reader():
rs = []
for r in readers:
rs.append(r())
for e in itertools.chain(*rs):
yield e
return reader
class ComposeNotAligned(ValueError):
pass
def compose(*readers, **kwargs):
"""
Creates a data reader whose output is the combination of input readers.
If input readers output following data entries:
(1, 2) 3 (4, 5)
The composed reader will output:
(1, 2, 3, 4, 5)
:param readers: readers that will be composed together.
:param check_alignment: if True, will check if input readers are aligned
correctly. If False, will not check alignment and trailing outputs
will be discarded. Defaults to True.
:type check_alignment: bool
:return: the new data reader.
:raises ComposeNotAligned: outputs of readers are not aligned.
Will not raise when check_alignment is set to False.
"""
check_alignment = kwargs.pop('check_alignment', True)
def make_tuple(x):
if isinstance(x, tuple):
return x
else:
return (x, )
def reader():
rs = []
for r in readers:
rs.append(r())
if not check_alignment:
for outputs in itertools.izip(*rs):
yield sum(map(make_tuple, outputs), ())
else:
for outputs in itertools.izip_longest(*rs):
for o in outputs:
if o is None:
# None will be not be present if compose is aligned
raise ComposeNotAligned(
"outputs of readers are not aligned.")
yield sum(map(make_tuple, outputs), ())
return reader
def buffered(reader, size):
"""
Creates a buffered data reader.
The buffered data reader will read and save data entries into a
buffer. Reading from the buffered data reader will proceed as long
as the buffer is not empty.
:param reader: the data reader to read from.
:type reader: callable
:param size: max buffer size.
:type size: int
:returns: the buffered data reader.
"""
class EndSignal():
pass
end = EndSignal()
def read_worker(r, q):
for d in r:
q.put(d)
q.put(end)
def data_reader():
r = reader()
q = Queue(maxsize=size)
t = Thread(
target=read_worker, args=(
r,
q, ))
t.daemon = True
t.start()
e = q.get()
while e != end:
yield e
e = q.get()
return data_reader
def firstn(reader, n):
"""
Limit the max number of samples that reader could return.
:param reader: the data reader to read from.
:type reader: callable
:param n: the max number of samples that return.
:type n: int
:return: the decorated reader.
:rtype: callable
"""
# TODO(yuyang18): Check if just drop the reader, could clean the opened
# resource or not?
def firstn_reader():
for i, item in enumerate(reader()):
if i == n:
break
yield item
return firstn_reader
class XmapEndSignal():
pass
def xmap_readers(mapper, reader, process_num, buffer_size, order=False):
"""
Use multiprocess to map samples from reader by a mapper defined by user.
And this function contains a buffered decorator.
:param mapper: a function to map sample.
:type mapper: callable
:param reader: the data reader to read from
:type reader: callable
:param process_num: process number to handle original sample
:type process_num: int
:param buffer_size: max buffer size
:type buffer_size: int
:param order: keep the order of reader
:type order: bool
:return: the decarated reader
:rtype: callable
"""
end = XmapEndSignal()
# define a worker to read samples from reader to in_queue
def read_worker(reader, in_queue):
for i in reader():
in_queue.put(i)
in_queue.put(end)
# define a worker to read samples from reader to in_queue with order flag
def order_read_worker(reader, in_queue):
in_order = 0
for i in reader():
in_queue.put((in_order, i))
in_order += 1
in_queue.put(end)
# define a worker to handle samples from in_queue by mapper
# and put mapped samples into out_queue
def handle_worker(in_queue, out_queue, mapper):
sample = in_queue.get()
while not isinstance(sample, XmapEndSignal):
r = mapper(sample)
out_queue.put(r)
sample = in_queue.get()
in_queue.put(end)
out_queue.put(end)
# define a worker to handle samples from in_queue by mapper
# and put mapped samples into out_queue by order
def order_handle_worker(in_queue, out_queue, mapper, out_order):
ins = in_queue.get()
while not isinstance(ins, XmapEndSignal):
order, sample = ins
r = mapper(sample)
while order != out_order[0]:
pass
out_queue.put(r)
out_order[0] += 1
ins = in_queue.get()
in_queue.put(end)
out_queue.put(end)
def xreader():
in_queue = Queue(buffer_size)
out_queue = Queue(buffer_size)
out_order = [0]
# start a read worker in a thread
target = order_read_worker if order else read_worker
t = Thread(target=target, args=(reader, in_queue))
t.daemon = True
t.start()
# start several handle_workers
target = order_handle_worker if order else handle_worker
args = (in_queue, out_queue, mapper, out_order) if order else (
in_queue, out_queue, mapper)
workers = []
for i in xrange(process_num):
worker = Thread(target=target, args=args)
worker.daemon = True
workers.append(worker)
for w in workers:
w.start()
sample = out_queue.get()
while not isinstance(sample, XmapEndSignal):
yield sample
sample = out_queue.get()
finish = 1
while finish < process_num:
sample = out_queue.get()
if isinstance(sample, XmapEndSignal):
finish += 1
else:
yield sample
return xreader
def _buf2lines(buf, line_break="\n"):
# FIXME: line_break should be automatically configured.
lines = buf.split(line_break)
return lines[:-1], lines[-1]
class PipeReader:
"""
PipeReader read data by stream from a command, take it's
stdout into a pipe buffer and redirect it to the parser to
parse, then yield data as your desired format.
You can using standard linux command or call another program
to read data, from HDFS, Ceph, URL, AWS S3 etc:
.. code-block:: python
cmd = "hadoop fs -cat /path/to/some/file"
cmd = "cat sample_file.tar.gz"
cmd = "curl http://someurl"
cmd = "python print_s3_bucket.py"
An example:
.. code-block:: python
def example_reader():
for f in myfiles:
pr = PipeReader("cat %s"%f)
for l in pr.get_line():
sample = l.split(" ")
yield sample
"""
def __init__(self, command, bufsize=8192, file_type="plain"):
if not isinstance(command, str):
raise TypeError("left_cmd must be a string")
if file_type == "gzip":
self.dec = zlib.decompressobj(
32 + zlib.MAX_WBITS) # offset 32 to skip the header
self.file_type = file_type
self.bufsize = bufsize
self.process = subprocess.Popen(
command.split(" "), bufsize=bufsize, stdout=subprocess.PIPE)
def get_line(self, cut_lines=True, line_break="\n"):
"""
:param cut_lines: cut buffer to lines
:type cut_lines: bool
:param line_break: line break of the file, like \n or \r
:type line_break: string
:return: one line or a buffer of bytes
:rtype: string
"""
remained = ""
while True:
buff = self.process.stdout.read(self.bufsize)
if buff:
if self.file_type == "gzip":
decomp_buff = self.dec.decompress(buff)
elif self.file_type == "plain":
decomp_buff = buff
else:
raise TypeError("file_type %s is not allowed" %
self.file_type)
if cut_lines:
lines, remained = _buf2lines(''.join(
[remained, decomp_buff]), line_break)
for line in lines:
yield line
else:
yield decomp_buff
else:
break
python/paddle/v2/reader/tests/CMakeLists.txt 0 → 100644
浏览文件 @ 1f6e0448
py_test(creator_test SRCS creator_test.py)
py_test(decorator_test SRCS decorator_test.py)
python/paddle/v2/reader/tests/__init__.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
python/paddle/v2/reader/tests/creator_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# Copyright PaddlePaddle contributors. All Rights Reservedd
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
import unittest
import numpy as np
import paddle.v2.reader.creator
class TestNumpyArray(unittest.TestCase):
def test_numpy_array(self):
l = [[1, 2, 3], [4, 5, 6]]
x = np.array(l, np.int32)
reader = paddle.v2.reader.creator.np_array(x)
for idx, e in enumerate(reader()):
self.assertItemsEqual(e, l[idx])
class TestTextFile(unittest.TestCase):
def test_text_file(self):
path = os.path.join(os.path.dirname(__file__), "test_data_creator.txt")
reader = paddle.v2.reader.creator.text_file(path)
for idx, e in enumerate(reader()):
self.assertEqual(e, str(idx * 2) + " " + str(idx * 2 + 1))
class TestRecordIO(unittest.TestCase):
def do_test(self, path):
reader = paddle.v2.reader.creator.recordio(path)
idx = 0
for e in reader():
if idx == 0:
self.assertEqual(e, (1, 2, 3))
elif idx == 1:
self.assertEqual(e, (4, 5, 6))
idx += 1
self.assertEqual(idx, 2)
def test_recordIO(self):
self.do_test(
os.path.join(
os.path.dirname(__file__), "test_reader_recordio.dat"))
self.do_test([
os.path.join(
os.path.dirname(__file__), "test_reader_recordio.dat")
])
if __name__ == '__main__':
unittest.main()
python/paddle/v2/reader/tests/decorator_test.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import time
import unittest
import paddle.v2.reader
def reader_creator_10(dur):
def reader():
for i in range(10):
# this invocation helps testing paddle.reader.buffer
time.sleep(dur)
yield i
return reader
class TestMap(unittest.TestCase):
def test_map(self):
d = {"h": 0, "i": 1}
def tokenize(x):
return d[x]
def read():
yield "h"
yield "i"
r = paddle.v2.reader.map_readers(tokenize, read)
for i, e in enumerate(r()):
self.assertEqual(e, i)
class TestBuffered(unittest.TestCase):
def test_read(self):
for size in range(20):
b = paddle.v2.reader.buffered(reader_creator_10(0), size)
c = 0
for i in b():
self.assertEqual(i, c)
c += 1
self.assertEqual(c, 10)
def test_buffering(self):
# read have 30ms delay.
b = paddle.v2.reader.buffered(reader_creator_10(0.03), 10)
last_time = time.time()
for idx, i in enumerate(b()):
elapsed_time = time.time() - last_time
if i == 0:
time.sleep(0.3)
else:
# read time should be short, meaning already buffered.
self.assertLess(elapsed_time, 0.05)
last_time = time.time()
class TestCompose(unittest.TestCase):
def test_compse(self):
reader = paddle.v2.reader.compose(
reader_creator_10(0), reader_creator_10(0))
for idx, e in enumerate(reader()):
self.assertEqual(e, (idx, idx))
def test_compose_not_aligned(self):
total = 0
reader = paddle.v2.reader.compose(
paddle.v2.reader.chain(reader_creator_10(0), reader_creator_10(0)),
reader_creator_10(0))
with self.assertRaises(paddle.v2.reader.ComposeNotAligned):
for e in reader():
total += 1
# expecting 10, not 20
self.assertEqual(total, 10)
def test_compose_not_aligned_no_check(self):
total = 0
reader = paddle.v2.reader.compose(
paddle.v2.reader.chain(reader_creator_10(0), reader_creator_10(0)),
reader_creator_10(0),
check_alignment=False)
for e in reader():
total += 1
# expecting 10, not 20
self.assertEqual(total, 10)
class TestChain(unittest.TestCase):
def test_chain(self):
c = paddle.v2.reader.chain(reader_creator_10(0), reader_creator_10(0))
idx = 0
for e in c():
self.assertEqual(e, idx % 10)
idx += 1
self.assertEqual(idx, 20)
class TestShuffle(unittest.TestCase):
def test_shuffle(self):
case = [(0, True), (1, True), (10, False), (100, False)]
a = reader_creator_10(0)
for size, checkEq in case:
s = paddle.v2.reader.shuffle(a, size)
total = 0
for idx, e in enumerate(s()):
if checkEq:
self.assertEqual(idx, e)
total += 1
self.assertEqual(total, 10)
class TestXmap(unittest.TestCase):
def test_xmap(self):
def mapper(x):
return (x + 1)
orders = (True, False)
thread_nums = (1, 2, 4, 8, 16)
buffered_size = (1, 2, 4, 8, 16)
for order in orders:
for tNum in thread_nums:
for size in buffered_size:
reader = paddle.v2.reader.xmap_readers(mapper,
reader_creator_10(0),
tNum, size, order)
for n in xrange(3):
result = []
for i in reader():
result.append(i)
if not order:
result.sort()
for idx, e in enumerate(result):
self.assertEqual(e, mapper(idx))
class TestPipeReader(unittest.TestCase):
def test_pipe_reader(self):
def example_reader(myfiles):
for f in myfiles:
pr = paddle.v2.reader.PipeReader("cat %s" % f, bufsize=128)
for l in pr.get_line():
yield l
import tempfile
records = [str(i) for i in xrange(5)]
temp = tempfile.NamedTemporaryFile()
try:
with open(temp.name, 'w') as f:
for r in records:
f.write('%s\n' % r)
result = []
for r in example_reader([temp.name]):
result.append(r)
for idx, e in enumerate(records):
self.assertEqual(e, result[idx])
finally:
# delete the temporary file
temp.close()
if __name__ == '__main__':
unittest.main()
python/paddle/v2/tests/CMakeLists.txt
浏览文件 @ 1f6e0448
py_test(test_op SRCS test_op.py) py_test(test_op SRCS test_op.py)
py_test(test_image SRCS test_image.py)
py_test(test_layer SRCS test_layer.py) py_test(test_layer SRCS test_layer.py)
py_test(test_topology SRCS test_topology.py) py_test(test_topology SRCS test_topology.py)
py_test(test_rnn_layer SRCS test_rnn_layer.py) py_test(test_rnn_layer SRCS test_rnn_layer.py)
......
python/paddle/v2/tests/test_image.py 0 → 100644
浏览文件 @ 1f6e0448
# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
import numpy as np
import paddle.v2.image as image
class Image(unittest.TestCase):
def test_resize_flip_chw(self):
# resize
im = image.load_image('cat.jpg')
im = image.resize_short(im, 256)
self.assertEqual(256, min(im.shape[:2]))
self.assertEqual(3, im.shape[2])
# flip
im = image.left_right_flip(im)
im2 = np.flip(im, 1)
self.assertEqual(im.all(), im2.all())
# to_chw
h, w, c = im.shape
im = image.to_chw(im)
self.assertEqual(c, im.shape[0])
self.assertEqual(h, im.shape[1])
self.assertEqual(w, im.shape[2])
if __name__ == '__main__':
unittest.main()
python/paddle/v2/tests/test_paramconf_order.py
浏览文件 @ 1f6e0448
...@@ -27,7 +27,6 @@ ...@@ -27,7 +27,6 @@
# limitations under the License. # limitations under the License.
import unittest import unittest
import math import math
import paddle.dataset as dataset
import paddle.v2 as paddle import paddle.v2 as paddle
...@@ -41,7 +40,7 @@ def wordemb(inlayer): ...@@ -41,7 +40,7 @@ def wordemb(inlayer):
def train(): def train():
word_dict = dataset.imikolov.build_dict() word_dict = paddle.dataset.imikolov.build_dict()
dict_size = len(word_dict) dict_size = len(word_dict)
# Every layer takes integer value of range [0, dict_size) # Every layer takes integer value of range [0, dict_size)
firstword = paddle.layer.data( firstword = paddle.layer.data(
......
python/setup.py.in
浏览文件 @ 1f6e0448
...@@ -77,6 +77,8 @@ if '${WITH_FLUID_ONLY}'== 'OFF': ...@@ -77,6 +77,8 @@ if '${WITH_FLUID_ONLY}'== 'OFF':
'paddle.v2', 'paddle.v2',
'paddle.v2.master', 'paddle.v2.master',
'paddle.v2.plot', 'paddle.v2.plot',
'paddle.v2.reader',
'paddle.v2.dataset',
'py_paddle'] 'py_paddle']
with open('@PADDLE_SOURCE_DIR@/python/requirements.txt') as f: with open('@PADDLE_SOURCE_DIR@/python/requirements.txt') as f:
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或