Skip to content

P
Paddle

PaddlePaddle / Paddle
1 年多 前同步成功

通知 2302
Star 20931
Fork 5422
P
Paddle
提交 19db989e 编写于 作者: W wanghaoshuang
浏览文件
操作

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

上级 61f4706b fb8c1cf0
展开全部 显示空白变更内容
内联 并排
Showing with 8344 addition and 590 deletion
benchmark/cluster/README.md
浏览文件 @ 19db989e
......@@ -36,11 +36,41 @@
- Trainer Count: 100
- Metrics: mini-batch / sec
| Batch Size | 32 | 64 | 128 | 256 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | - | - | - | - |
| PaddlePaddle v2 | - | - | - | - |
| TensorFlow | - | - | - | - |
<table>
<thead>
<tr>
<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
......@@ -48,11 +78,41 @@
- Batch Size: 64
- Metrics: mini-batch / sec
| PServer Count | 10 | 20 | 40 | 60 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | - | - | - | - |
| PaddlePaddle v2 | - | - | - | - |
| TensorFlow | - | - | - | - |
<table>
<thead>
<tr>
<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
......@@ -67,11 +127,69 @@ The parallel efficiency is:
$E = \div(S, N)$
| Trainer Counter | 1 | 10 | 20 | 30 | 40 | 50 | 60 | 70 | 80 | 90 | 100 |
| -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | - | - | - | - | - | - | - | - | - | - | - |
| PaddlePaddle v2 | - | - | - | - | - | - | - | - | - | - | - | - |
| TensorFlow | - | - | - | - | - | - | - | - | - | - | - | - | - |
<table>
<thead>
<tr>
<th>Trainer Counter </th>
<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
......
benchmark/cluster/vgg16/README.md
浏览文件 @ 19db989e
......@@ -16,11 +16,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Metrics: samples / sec
| Batch Size | 32 | 64 | 128 | 256 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | 15.44 | 16.32 | 16.74 | 16.79 |
| PaddlePaddle v2 | 15.97 | 17.04 | 17.60 | 17.83 |
| TensorFlow | 9.09 | 9.10 | 9.24 | 8.66 |
<table>
<thead>
<tr>
<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> 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
......@@ -28,12 +58,40 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Trainer Count: 20
- Metrics: samples / sec
| Batch Size | 32 | 64 | 128 | 256 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | 190.20 | 222.15 | 247.40 | 258.18 |
| PaddlePaddle v2 | 170.96 | 233.71 | 256.14 | 329.23 |
| TensorFlow | - | - | - | - |
<table>
<thead>
<tr>
<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> 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
......@@ -41,11 +99,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Batch Size: 128
- Metrics: samples / sec
| Trainer Count | 20 | 40 | 80 | 100 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid | 263.29 (78.64%) | 518.80 (77.47%) | 836.26 (62.44%) | 1019.29 (60.89%) |
| PaddlePaddle v2 (need more tests) | 326.85 (92.85%) | 534.58 (75.93%) | 853.30 (60.60%) | 1041.99 (59.20%) |
| TensorFlow | - | - | - | - |
<table>
<thead>
<tr>
<th>Trainer Count </th>
<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
......@@ -53,11 +141,41 @@ Setting environment variable: `MKL_NUM_THREADS=1`.
- Batch Size: 128
- Metrics: samples/ sec
| PServer Count | 3 | 6 |10 | 20 |
| -- | -- | -- | -- | -- |
| PaddlePaddle Fluid(should fix in next PR) | 589.1 | 592.6 | 656.4 | 655.8 |
| PaddlePaddle v2 | 593.4 | 791.3 | 729.7 | 821.7 |
| TensorFlow | - | - | - | - |
<table>
<thead>
<tr>
<th>PServer Count </th>
<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.*
......
benchmark/fluid/machine_translation.py 0 → 100644
浏览文件 @ 19db989e
# 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 @ 19db989e
# 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 @ 19db989e
# 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 @ 19db989e
#!/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 @ 19db989e
# 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 @ 19db989e
# 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 @ 19db989e
......@@ -494,6 +494,12 @@ reshape
.. autofunction:: paddle.fluid.layers.reshape
:noindex:
pad
---
.. autofunction:: paddle.fluid.layers.pad
:noindex:
scale
-----
......
doc/fluid/design/algorithm/parameter_average.md
浏览文件 @ 19db989e
......@@ -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:
<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.
......
doc/fluid/design/concepts/README.md
浏览文件 @ 19db989e
......@@ -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:
| C++ | CUDA C++ | Go |
|---|---|---|
| cc_library | nv_library | go_library |
| cc_binary | nv_binary | go_binary |
| cc_test | nv_test | go_test |
<table>
<thead>
<tr>
<th>C++</th>
<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 `_binary` functions generate executable binary files.
......
doc/fluid/design/concepts/block.md
浏览文件 @ 19db989e
......@@ -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:
| programming languages | PaddlePaddle |
|-----------------------|-----------------------|
| for, while loop | RNN, WhileOp |
| if, if-else, switch | IfElseOp, SwitchOp |
| sequential execution | a sequence of layers |
<table>
<thead>
<tr>
<th>programming languages</th>
<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.
......@@ -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:
| programming languages | PaddlePaddle |
|-----------------------|---------------------------------|
| stack | scope hierarchy |
| stack frame | scope |
| push at entering block| push at entering block |
| pop at leaving block | destroy when minibatch completes|
<table>
<thead>
<tr>
<th>programming languages</th>
<th>PaddlePaddle</th>
</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:
......
doc/fluid/design/concepts/functions_operators_layers.md
浏览文件 @ 19db989e
......@@ -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:
| C++ functions/functors | mul | add | | |
|------------------------|--------------|--------------|-------------|----------|
| C++ operator class | mulOp | addOp | FCOp | |
| Python binding | operator.mul | operator.add | operator.fc | |
| Python function | | | | layer.fc |
<table>
<thead>
<tr>
<th>C++ functions/functors</th>
<th>mul</th>
<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:
......
doc/fluid/design/concepts/lod_tensor.md
浏览文件 @ 19db989e
......@@ -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.
| | TensorFlow | PaddlePaddle |
|-----------------------|------------|--------------|
| RNN | Support | Support |
| recursive RNN | Support | Support |
| padding zeros | Must | No need |
| blob data type | Tensor | LoDTensor |
<table>
<thead>
<tr>
<th></th>
<th>TensorFlow</th>
<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.
......
doc/fluid/design/concepts/var_desc.md
浏览文件 @ 19db989e
......@@ -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.
| |compile time|runtime|
|---|---|---|
|Data|VarDesc(proto)|Variable(cpp)|
|Operation|OpDesc(proto)|Operator(cpp)|
<table>
<thead>
<tr>
<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
......
doc/fluid/design/concurrent/concurrent_programming.md
浏览文件 @ 19db989e
......@@ -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
| Go | Fluid |
|----|-------|
|user-defined functions | [layers](https://github.com/PaddlePaddle/Paddle/tree/develop/python/paddle/fluid) |
| control-flow and built-in functions | [intrinsics/operators](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/operators) |
| goroutines, channels | [class ThreadPool](https://github.com/PaddlePaddle/Paddle/tree/develop/paddle/framework/thread_pool.h) |
| runtime | [class Executor](https://github.com/PaddlePaddle/Paddle/blob/develop/paddle/framework/executor.h) |
<table>
<thead>
<tr>
<th></th>
<th>Go</th>
<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
......
doc/fluid/design/concurrent/csp.md
浏览文件 @ 19db989e
......@@ -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:
| concurrent programming model | implementation |
|-----|-----|
| mutex | types and functions in standard libraries |
| semaphore | types and functions in standard libraries |
| communicating sequential processes (CSP) | Go programming language |
| actor model | Erlang programming language |
| message passing | MPI |
| bulk synchronous parallel (BSP) | Pregel distributed programming framework |
<table>
<thead>
<tr>
<th>concurrent programming model</th>
<th>implementation</th>
</tr>
</thead>
<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.
......
doc/fluid/design/modules/python_api.md
浏览文件 @ 19db989e
......@@ -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.
| Python classes | Protobuf messages |
| --- | --- |
| Program | ProgramDesc |
| Block | BlockDesc |
| Operator | OpDesc |
| Variable | VarDesc |
<table>
<thead>
<tr>
<th>Python classes</th>
<th>Protobuf messages</th>
</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.
......
doc/fluid/design/motivation/fluid.md
浏览文件 @ 19db989e
......@@ -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.
| Existed since | model as sequence of layers | model as graph of operators | No model |
|--|--|--|--|
| 2013 | Caffe, Theano, Torch, PaddlePaddle | | |
| 2015 | | TensorFlow, MxNet, Caffe2, ONNX, n-graph | |
| 2016 | | | PyTorch, TensorFlow Eager Execution, PaddlePaddle Fluid |
<table>
<thead>
<tr>
<th>Existed since</th>
<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.
......
doc/fluid/design/motivation/refactorization.md
浏览文件 @ 19db989e
......@@ -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.
| | Representation (protobuf messages) | Realization (C++ class objects) |
|---|---|---|
|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)|
|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)|
|Block|BlockDesc|Block|
<table>
<thead>
<tr>
<th></th>
<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 `}`).
......
doc/fluid/design/network/deep_speech_2.md
浏览文件 @ 19db989e
......@@ -68,11 +68,33 @@ We roughly break down the project into 14 tasks:
Tasks parallelizable within phases:
Roadmap | Description | Parallelizable Tasks
----------- | :------------------------------------ | :--------------------
Phase I | Simplified model & components | *Task 1* ~ *Task 8*
Phase II | Standard model & benchmarking & profiling | *Task 9* ~ *Task 12*
Phase III | Documentations | *Task13* ~ *Task14*
<table>
<thead>
<tr>
<th>Roadmap</th>
<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!
......@@ -121,18 +143,63 @@ Key ingredients about the layers:
- 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.
<table>
<thead>
<tr>
<th>Required Components</th>
<th> PaddlePaddle Support</th>
<th> Need to Develop</th>
</tr>
</thead>
<tbody>
<tr>
<td>Data Layer I (Spectrogram) </td>
<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>
Required Components | PaddlePaddle Support | Need to Develop
:------------------------------------- | :-------------------------------------- | :-----------------------
Data Layer I (Spectrogram) | Not supported yet. | TBD (Task 3)
Data Layer II (Transcription) | `paddle.data_type.integer_value_sequence` | -
2D Convolution Layer | `paddle.layer.image_conv_layer` | -
DataType Converter (vec2seq) | `paddle.layer.block_expand` | -
Bi-/Uni-directional RNNs | `paddle.layer.recurrent_group` | -
Row Convolution Layer | Not supported yet. | TBD (Task 4)
CTC-loss Layer | `paddle.layer.warp_ctc` | -
Batch Normalization Layer | `paddle.layer.batch_norm` | -
CTC-Beam search | Not supported yet. | TBD (Task 6)
### Row Convolution
......
doc/fluid/dev/index_cn.rst
浏览文件 @ 19db989e
......@@ -4,9 +4,9 @@
.. toctree::
:maxdepth: 1
new_op_en.md
new_op_kernel_en.md
use_eigen_en.md
new_op_cn.md
new_op_kernel.md
use_eigen_cn.md
name_convention.md
support_new_device.md
releasing_process.md
......
doc/fluid/dev/index_en.rst
浏览文件 @ 19db989e
......@@ -5,7 +5,7 @@ Development
:maxdepth: 1
new_op_en.md
new_op_kernel_en.md
new_op_kernel.md
use_eigen_en.md
name_convention.md
support_new_device.md
......
doc/fluid/dev/new_op_cn.md
浏览文件 @ 19db989e
......@@ -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：
内容 | 定义位置
-------------- | :----------------------
OpProtoMake定义 | `.cc`文件，Backward Op不需要定义OpProtoMake
Op定义 | `.cc`文件
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。
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
<table>
<thead>
<tr>
<th>内容</th>
<th>定义位置</th>
</tr>
</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。**
......
doc/fluid/dev/new_op_en.md
浏览文件 @ 19db989e
......@@ -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.
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.**
......
doc/fluid/dev/releasing_process.md
浏览文件 @ 19db989e
......@@ -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。
| | 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 |
| --- | --- | --- | --- | --- | --- | --- | --- | --- |
| API.V2 + Docker + GPU | | | | | | | | |
| API.V2 + Docker + CPU | | | | | | | | |
| `paddle_trainer` + Docker + GPU | | | | | | | | |
| `paddle_trainer` + Docker + CPU | | | | | | | | |
| API.V2 + Ubuntu + GPU | | | | | | | | |
| API.V2 + Ubuntu + CPU | | | | | | | | |
| `paddle_trainer` + Ubuntu + GPU | | | | | | | | |
| `paddle_trainer` + Ubuntu + CPU | | | | | | | | |
<table>
<thead>
<tr>
<th></th>
<th>新手入门章节 </th>
<th> 识别数字</th>
<th> 图像分类</th>
<th>词向量</th>
<th> 情感分析</th>
<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 @ 19db989e
......@@ -16,18 +16,58 @@ As a result, we design a particular format for tensor serialization. By default,
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 |
| --- | --- | --- |
| version | uint32_t | Version of saved file. Always 0 now. |
| tensor desc length | uint32_t | TensorDesc(Protobuf message) length in bytes. |
| tensor desc | void* | TensorDesc protobuf binary message |
| tensor data | void* | Tensor's data in binary format. The length of `tensor_data` is decided by `TensorDesc.dims()` and `TensorDesc.data_type()` |
| lod_level | uint64_t | Level of LoD |
| length of lod[0] | uint64_t | [Optional] length of lod[0] in bytes. |
| data of lod[0] | uint64_t* | [Optional] lod[0].data() |
| ... | ... | ... |
<table>
<thead>
<tr>
<th>field name</th>
<th>type </th>
<th>description </th>
</tr>
</thead>
<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
......
doc/fluid/howto/cluster/fluid_cluster_train_cn.md
浏览文件 @ 19db989e
......@@ -99,13 +99,49 @@ for pass_id in range(100):
### 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:
| 参数名 | 值类型 | 说明 | 示例 |
|:-------------|:------|:---------------------------------------|:-------------|
| 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 |
| pservers | str | parameter server 列表 | 127.0.0.1:6710,127.0.0.1:6711 |
| trainers | int | 训练节点的总个数，>0的数字 | 4 |
| server_endpoint | str | 当前所起的服务节点的IP:PORT | 127.0.0.1:8789 |
| training_role | str | 节点角色， TRAINER/PSERVER | PSERVER |
<table>
<thead>
<tr>
<th>参数名</th>
<th> 值类型</th>
<th>说明</th>
<th> 示例</th>
</tr>
</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：
......
doc/fluid/howto/optimization/cpu_profiling_cn.md
浏览文件 @ 19db989e
......@@ -42,14 +42,40 @@ cprofilev -a 0.0.0.0 -p 3214 -f profile.out main.py
每一列的含义是:
| 列名 | 含义 |
| --- | --- |
| ncalls | 函数的调用次数 |
| 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 |
| percall | tottime的每次调用平均时间 |
| cumtime | 函数总时间。包含这个函数调用其他函数的时间 |
| percall | cumtime的每次调用平均时间 |
| filename:lineno(function) | 文件名, 行号，函数名 |
<table>
<thead>
<tr>
<th>列名</th>
<th>含义 </th>
</tr>
</thead>
<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
......
doc/fluid/howto/optimization/cpu_profiling_en.md
浏览文件 @ 19db989e
......@@ -57,14 +57,40 @@ port, we will see the output like the following:
where each line corresponds to Python function, and the meaning of
each column is as follows:
| column | meaning |
| --- | --- |
| ncalls | the number of calls into a function |
| tottime | the total execution time of the function, not including the execution time of other functions called by the function |
| percall | tottime divided by ncalls |
| cumtime | the total execution time of the function, including the execution time of other functions being called |
| percall | cumtime divided by ncalls |
| filename:lineno(function) | where the function is defined |
<table>
<thead>
<tr>
<th>column</th>
<th>meaning </th>
</tr>
</thead>
<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
......
doc/v2/faq/build_and_install/index_en.rst
浏览文件 @ 19db989e
############################
Install, Build and Unit test
############################
.. _install_faq:
###############################
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
TBD
doc/v2/howto/rnn/recurrent_group_en.md
浏览文件 @ 19db989e
# Recurrent Group Tutorial
TBD
## Overview
Sequential data is common in natural language processing.
A sentence is a sequence of words and many sentences form a paragraph further. Therefore, a paragraph can be viewed as a nested sequence with two level, where each element of the sequence is another sequence. That is to say, sequential data could be recursive. An example of two-level recursive sequential data is that an article is composed of a sequence of sentences, and each sentence a sequence of words.
PaddlePaddle and PaddlePaddle v2 support two-level recursive sequential data. The two-level sequence is a very flexible data, which helps us to better describe more complex language data such as discribing paragraphs and several rounds of dialogues. Based on two-level sequence input, we can design and build a flexible, hierarchical RNN model that encodes input data from the word and sentence level. For the support of arbitrary levels, please refer to PaddlePaddle Fluid.
In PaddlePaddle, `recurrent_group` is an arbitrarily complex RNN unit. The user only needs to define the calculation that the RNN will complete in one time step. PaddlePaddle is responsible for the propagation of information and error in time series.
Furthermore, `recurrent_group` can also be extended to handle two-level sequence. By defining two nested `recurrent_group` operations at the clause level and the word level respectively, a hierarchical and complex RNN is finally achieved.
Currently, in the PaddlePaddle, there are `recurrent_group` and some Layers that can process bidirectional sequences. For details, refer to the document: <a href = "hierarchical_layer_en.html">Layers for supporting double-layer sequences as input.</a>
## Related Concepts
### Basic Principle
`recurrent_group` is an arbitrarily complex RNN unit supported by PaddlePaddle. The user only needs to focus on the calculations that the RNN is designed to complete within a single time step. The PaddlePaddle is responsible for completing the propagation of information and gradients over time.
In PaddlePaddle, a simple call to `recurrent_group` is as follows:
``` python
recurrent_group(step, input, reverse)
```
- step: A callable function that defines the calculations completed by the RNN unit within a time step
- input: The input must be a single-layer sequence or a double-layer sequence
- reverse: Whether to process the input sequence in reverse order
The core of using `recurrent_group` is to design the logic of the step function. The step function can be freely combined with various layers supported by PaddlePaddle to complete arbitrary arithmetic logic. The input of `recurrent_group` (input) becomes the input of the step function. Since the step function only focuses on the calculation within one time step of RNN, here `recurrent_group` completes the splitting of the original input data for us.
### Input
The input sequence processed by `recurrent_group` is mainly divided into the following three types:
- **Input Data**: When putting a two-level sequence into `recurrent_group`, it will be disassembled into a single-level sequence. When putting a single-level sequence into `recurrent_group`, it will be disassembled into a non-sequence and then passed to the step function. This process is completely transparent to the user. There are two possible types: 1) User input via data_layer; 2) Output from other layers.
- **Read-only Memory Input**: `StaticInput` defines a read-only Memory. The input specified by `StaticInput` will not be disassembled by `recurrent_group`, and each time step of the `recurrent_group` loop will always be able to reference all inputs. It may be a non-sequence or a single-layer sequence.
- **Input of Sequence Generation Task**: `GeneratedInput` is only used to specify input data in a sequence generation task.
### Input Example
Sequence generation tasks mostly follow the encoder-decoer architecture. The encoder and decoder can be arbitrary neural network units capable of processing sequences and RNN is the most popular choice.
Given the encoder output and the current word, the decoder predicts the next most likely word each time. In this structure, the decoder accepts two inputs:
- Target sequence to be generated: a input of the decoder and the basis of the decoder loop. `recurrent_group` will disassemble this input type.
- Encoder output, an non-sequencce or single-sequence: a unbounded memory. Each time step in the decoder loop will reference the entire result and should not be disassembled. This type of input must be specified via `StaticInput`. For more discussion on Unbounded Memory, please refer to the paper [Neural Turning Machine](https://arxiv.org/abs/1410.5401).
In a sequence generation task, the decoder RNN always refers to the word vector of the word predicted at the previous moment as the current time input. `GeneratedInput` will automate this process.
### Output
The `step` function must return the output of one or more Layers. The output of this Layer will be the final output of the entire `recurrent_group`. In the output process, `recurrent_group` will concatenate the output of each time step, which is also transparent to the user.
### Memory
Memory can only be defined and used in `recurrent_group`. Memory cannot exist independently and must point to a layer defined by PaddlePaddle. Memory is referenced to get a momentary output from this layer, so memory can be interpreted as a delay operation.
The user can explicitly specify the output of a layer to initialize the memory. When not specified, memory is initialized to 0 by default.
## Sequence-level RNN Introduction
`recurrent_group` helps us to split the input sequence, merge the output, and loop through the sequence of computational logic.
Using this feature, the two nested `recurrent_group` can handle the nested two-level sequences, implementing sequence-level RNN structures at both the word and sentence levels.
- Word-level RNN: each state corresponds to a word.
- Sequence-level RNN: a sequence-layer RNN consists of multiple word-layer RNNs. Each word-layer RNN (ie, each state of a sequence-layer RNN) has a subsequence.
For convenience of description, the following takes the NLP task as an example. A paragraph containing a subsequence is defined as a two-level sequence, and a sentence containing a word is defined as a single-layer sequence. Then, the zero-level sequence is a word.
## Usage of Sequence-level RNN
### Usage of Training Process
Using `recurrent_group` requires the following conventions:
- **Single-input Single-output**: Both input and output are single layer sequences.
- If there are multiple inputs, the number of words in different input sequences must be exactly equal.
- A single-layer sequence is output, and the number of words in the output sequence is the same as the input sequence.
- memory: define memory to point to a layer in the step function, get a moment output from this layer by referencing memory to form a recurrent connection. The is_seq parameter of memory must be false. If memory is not defined, the operations within each time step are independent.
- boot_layer: the initial state of memory, set 0 by default. is_seq in memory must be false.
- **Double-input Double-output**: Both input and output are two-level sequence.
- If there are multiple input sequences, the number of subsequence contained in different inputs must be strictly equal, but the number of words in the subsequence may not be equal.
- output a two-level sequence. The number of subsequence and the number of words are the same as the specified input sequence and the first input is default.
- memory: defining memory in the step function, pointing to a layer, by referring to the memory to get the output of this layer at a time, forming a recurrent connection. The memory defined in the outer `recurrent_group` step function can record the state of the previous subsequence, either as a single-level sequence (only as read-only memory) or as a word. If memory is not defined, the operations between subsequence are independent.
- boot_layer: the initial state of memory. It is either a single-level sequence (only as read-only memory) or a vector. The default is not set, that is, the initial state is 0.
- **Double-input Single-output**: not support for now, and output the error with "In hierachical RNN, all out links should be from sequences now".
### Usage of Generation Process
Using `beam_search` need follow those conventions:
- Word-level RNN: generate the next word from a word.
- Sequence-level RNN: the single-layer RNN generated subsequence is concatenated into a new double-layer sequence. Semantically, there is no case where a subsequence generates the next subseq directly.
paddle/fluid/framework/details/multi_devices_graph_builder.cc
浏览文件 @ 19db989e
......@@ -55,6 +55,7 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build(
const ProgramDesc &program) const {
auto graph = new SSAGraph();
SSAGraph &result = *graph;
std::unordered_set<std::string> og_has_been_broadcast;
result.vars_.resize(places_.size());
bool is_forwarding = true;
......@@ -122,9 +123,15 @@ std::unique_ptr<SSAGraph> MultiDevSSAGraphBuilder::Build(
if (!is_forwarding) {
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) {
if (grad_names_.count(og) != 0) { // is param grad
if (grad_names_.count(og) != 0 &&
og_has_been_broadcast.count(og) == 0) { // is param grad
// Insert NCCL AllReduce Op
og_has_been_broadcast.insert(og);
#ifdef PADDLE_WITH_CUDA
result.ops_.emplace_back(
new NCCLAllReduceOpHandle(local_scopes_, places_, *nccl_ctxs_));
......
paddle/fluid/framework/details/nccl_all_reduce_op_handle.cc
浏览文件 @ 19db989e
......@@ -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 framework
} // namespace paddle
paddle/fluid/framework/details/nccl_all_reduce_op_handle.h
浏览文件 @ 19db989e
......@@ -14,6 +14,9 @@
#pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/details/op_handle_base.h"
#include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/scope.h"
......@@ -34,6 +37,10 @@ struct NCCLAllReduceOpHandle : public OpHandleBase {
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:
void RunImpl() override;
};
......
paddle/fluid/framework/details/op_handle_base.h
浏览文件 @ 19db989e
......@@ -13,6 +13,8 @@
// limitations under the License.
#pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/details/var_handle.h"
#include "paddle/fluid/platform/device_context.h"
......@@ -53,6 +55,10 @@ class OpHandleBase {
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:
virtual void RunImpl() = 0;
};
......
paddle/fluid/framework/details/threaded_ssa_graph_executor.cc
浏览文件 @ 19db989e
......@@ -23,22 +23,36 @@ ThreadedSSAGraphExecutor::ThreadedSSAGraphExecutor(
size_t num_threads, bool use_event,
const std::vector<Scope *> &local_scopes,
const std::vector<platform::Place> &places,
std::unique_ptr<SSAGraph> &&graph)
std::unique_ptr<SSAGraph> &&graph, bool allow_op_delay)
: SSAGraphExecutor(std::move(graph)),
pool_(num_threads >= 2 ? new ::ThreadPool(num_threads) : nullptr),
local_scopes_(local_scopes),
places_(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(
const std::vector<std::string> &fetch_tensors) {
std::unordered_map<OpHandleBase *, size_t> pending_ops;
std::unordered_set<VarHandleBase *> pending_vars;
BlockingQueue<VarHandleBase *> ready_vars;
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) {
pending_vars.insert(&var);
......@@ -106,7 +120,14 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
auto run_all_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();
};
......@@ -118,13 +139,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
}
// Step 3. Execution
while (!pending_vars.empty()) {
while (!pending_vars.empty() || !ready_ops.empty() || !delayed_ops.empty()) {
// 1. Run All Ready ops
run_all_ready_ops();
// 2. Find ready variable
bool timeout;
auto cur_ready_vars = ready_vars.PopAll(1000, &timeout);
auto cur_ready_vars = ready_vars.PopAll(1, &timeout);
if (timeout) {
if (exception_) {
......@@ -141,13 +162,29 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
auto &deps = pending_ops[op];
--deps;
if (deps == 0) {
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.
}
PADDLE_ENFORCE(ready_ops.empty());
PADDLE_ENFORCE(delayed_ops.empty());
PADDLE_ENFORCE(blocked_by_delayed_ops.empty());
++computation_count_;
auto sync_computation = [&] {
......@@ -182,12 +219,13 @@ FeedFetchList ThreadedSSAGraphExecutor::Run(
}
void ThreadedSSAGraphExecutor::RunOp(
BlockingQueue<VarHandleBase *> &ready_var_q, details::OpHandleBase *op) {
auto op_run = [&ready_var_q, op, this] {
BlockingQueue<VarHandleBase *> *ready_var_q, details::OpHandleBase *op) {
auto op_run = [ready_var_q, op, this] {
try {
VLOG(10) << op->Name() << " : " << op->DebugString();
op->Run(use_event_);
ready_var_q.Extend(op->outputs_);
running_ops_--;
ready_var_q->Extend(op->outputs_);
} catch (platform::EnforceNotMet ex) {
exception_.reset(new platform::EnforceNotMet(ex));
} catch (...) {
......
paddle/fluid/framework/details/threaded_ssa_graph_executor.h
浏览文件 @ 19db989e
......@@ -14,7 +14,12 @@
#pragma once
#include <chrono>
#include <deque>
#include <string>
#include <unordered_set>
#include <utility>
#include <vector>
#include <functional>
#include "ThreadPool.h" // ThreadPool in thrird party
#include "paddle/fluid/framework/details/ssa_graph_executor.h"
......@@ -70,7 +75,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
ThreadedSSAGraphExecutor(size_t num_threads, bool use_event,
const std::vector<Scope *> &local_scopes,
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
// Use topological sort algorithm
......@@ -79,9 +85,11 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
~ThreadedSSAGraphExecutor() {}
private:
void RunOp(BlockingQueue<VarHandleBase *> &ready_var_q,
void RunOp(BlockingQueue<VarHandleBase *> *ready_var_q,
details::OpHandleBase *op);
void RunDelayedOps(const std::unordered_set<OpHandleBase *> &delayed_ops);
private:
std::unique_ptr<::ThreadPool> pool_;
std::vector<Scope *> local_scopes_;
......@@ -89,6 +97,8 @@ class ThreadedSSAGraphExecutor : public SSAGraphExecutor {
platform::DeviceContextPool fetch_ctxs_;
const bool use_event_;
std::unique_ptr<platform::EnforceNotMet> exception_;
std::atomic<int> running_ops_;
bool allow_op_delay_;
size_t computation_count_{0};
size_t max_async_computation{100};
......
paddle/fluid/framework/details/var_handle.h
浏览文件 @ 19db989e
......@@ -22,7 +22,7 @@
namespace paddle {
namespace framework {
namespace details {
struct OpHandleBase;
class OpHandleBase;
// VarHandleBase is the var node in the dependency graph.
// A variable can only be generated by a single operator. i.e.
......
paddle/fluid/framework/executor.cc
浏览文件 @ 19db989e
......@@ -279,6 +279,21 @@ std::unique_ptr<ExecutorPrepareContext> Executor::Prepare(
return std::unique_ptr<ExecutorPrepareContext>(ctx);
}
std::vector<std::shared_ptr<ExecutorPrepareContext>> Executor::Prepare(
const ProgramDesc& program, const std::vector<int>& block_ids) {
std::vector<std::shared_ptr<ExecutorPrepareContext>> result;
for (auto& bid : block_ids) {
auto* ctx = new ExecutorPrepareContext(program, bid);
PADDLE_ENFORCE_LT(static_cast<size_t>(bid), program.Size());
auto& block = program.Block(bid);
for (auto& op_desc : block.AllOps()) {
ctx->ops_.push_back(OpRegistry::CreateOp(*op_desc));
}
result.push_back(std::shared_ptr<ExecutorPrepareContext>(ctx));
}
return result;
}
void Executor::RunPreparedContext(ExecutorPrepareContext* ctx, Scope* scope,
bool create_local_scope, bool create_vars) {
auto& block = ctx->prog_.Block(ctx->block_id_);
......
paddle/fluid/framework/executor.h
浏览文件 @ 19db989e
......@@ -61,6 +61,9 @@ class Executor {
static std::unique_ptr<ExecutorPrepareContext> Prepare(
const ProgramDesc& program, int block_id);
static std::vector<std::shared_ptr<ExecutorPrepareContext>> Prepare(
const ProgramDesc& program, const std::vector<int>& block_ids);
void RunPreparedContext(ExecutorPrepareContext* ctx, Scope* scope,
bool create_local_scope = true,
bool create_vars = true);
......
paddle/fluid/framework/parallel_executor.cc
浏览文件 @ 19db989e
......@@ -13,9 +13,10 @@ See the License for the specific language governing permissions and
limitations under the License. */
#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
#include "paddle/fluid/platform/nccl_helper.h"
......@@ -47,7 +48,7 @@ ParallelExecutor::ParallelExecutor(
const std::vector<platform::Place> &places,
const std::unordered_set<std::string> &params,
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_->global_scope_ = scope;
......@@ -82,8 +83,8 @@ ParallelExecutor::ParallelExecutor(
auto graph = builder.Build(main_program);
member_->executor_.reset(new details::ThreadedSSAGraphExecutor(
num_threads, use_event, member_->local_scopes_, places,
std::move(graph)));
num_threads, use_event, member_->local_scopes_, places, std::move(graph),
allow_op_delay));
// Step 3. Create vars in each scope;
for (auto *scope : member_->local_scopes_) {
......@@ -151,6 +152,7 @@ void ParallelExecutor::BCastParamsToGPUs(
void ParallelExecutor::Run(const std::vector<std::string> &fetch_tensors,
const std::string &fetched_var_name) {
platform::RecordBlock b(0);
auto fetch_data = member_->executor_->Run(fetch_tensors);
*member_->global_scope_->Var(fetched_var_name)->GetMutable<FeedFetchList>() =
fetch_data;
......
paddle/fluid/framework/parallel_executor.h
浏览文件 @ 19db989e
......@@ -14,8 +14,9 @@ limitations under the License. */
#pragma once
#include <future>
#include <string>
#include <unordered_set>
#include <vector>
#include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/op_info.h"
#include "paddle/fluid/framework/program_desc.h"
......@@ -37,7 +38,8 @@ class ParallelExecutor {
const std::unordered_set<std::string>& params,
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);
void Run(const std::vector<std::string>& fetch_tensors,
const std::string& fetched_var_name = "fetched_var");
......
paddle/fluid/framework/selected_rows.h
浏览文件 @ 19db989e
......@@ -10,6 +10,9 @@ See the License for the specific language governing permissions and
limitations under the License. */
#pragma once
#include <vector>
#include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/tensor.h"
......@@ -52,7 +55,7 @@ class SelectedRows {
private:
// 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.
Vector<int64_t> rows_;
std::unique_ptr<Tensor> value_{nullptr};
......
paddle/fluid/operators/detail/CMakeLists.txt
浏览文件 @ 19db989e
......@@ -2,7 +2,7 @@ if(WITH_DISTRIBUTE)
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)
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 grpc_server_test 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
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)
......
paddle/fluid/operators/detail/grpc_client.cc
浏览文件 @ 19db989e
......@@ -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
limitations under the License. */
#include "grpc_client.h"
#include <sys/time.h>
#include "paddle/fluid/operators/detail/grpc_client.h"
#include <limits>
#include "paddle/fluid/framework/threadpool.h"
namespace paddle {
......@@ -52,7 +54,7 @@ bool RPCClient::AsyncSendVariable(const std::string& ep,
auto call = s->stub_g_.PrepareUnaryCall(
s->context_.get(), "/sendrecv.SendRecvService/SendVariable", req, &cq_);
call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s);
call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
});
req_count_++;
......@@ -70,8 +72,7 @@ void ProcGetResponse(const VarHandle& var_h,
template <typename T>
void RequestToByteBuffer(const T& proto, ::grpc::ByteBuffer* result) {
::grpc::Slice slice(proto.ByteSizeLong());
proto.SerializeWithCachedSizesToArray(
const_cast<uint8_t*>(reinterpret_cast<const uint8_t*>(slice.begin())));
proto.SerializeWithCachedSizesToArray(const_cast<uint8_t*>(slice.begin()));
::grpc::ByteBuffer tmp(&slice, 1);
result->Swap(&tmp);
}
......@@ -109,7 +110,7 @@ bool RPCClient::AsyncGetVariable(const std::string& ep,
auto call = s->stub_g_.PrepareUnaryCall(
s->context_.get(), "/sendrecv.SendRecvService/GetVariable", buf, &cq_);
call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s);
call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
});
req_count_++;
......@@ -153,7 +154,7 @@ bool RPCClient::AsyncPrefetchVariable(const std::string& ep,
s->context_.get(), "/sendrecv.SendRecvService/PrefetchVariable", req,
&cq_);
call->StartCall();
call->Finish(&s->reply_, &s->status_, (void*)s);
call->Finish(&s->reply_, &s->status_, static_cast<void*>(s));
});
req_count_++;
......@@ -169,7 +170,7 @@ void RPCClient::AsyncSendBatchBarrier(const std::string& ep, int64_t time_out) {
sendrecv::VariableMessage req;
req.set_varname(BATCH_BARRIER_MESSAGE);
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_++;
}
......@@ -181,7 +182,7 @@ void RPCClient::AsyncSendFetchBarrier(const std::string& ep, int64_t time_out) {
sendrecv::VariableMessage req;
req.set_varname(FETCH_BARRIER_MESSAGE);
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_++;
}
......
paddle/fluid/operators/detail/grpc_server.cc
浏览文件 @ 19db989e
......@@ -14,6 +14,9 @@ limitations under the License. */
#include "paddle/fluid/operators/detail/grpc_server.h"
#include <limits>
#include <string>
using ::grpc::ServerAsyncResponseWriter;
namespace paddle {
......@@ -156,6 +159,8 @@ class RequestPrefetch final : public RequestBase {
::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;
}
......@@ -221,6 +226,7 @@ void AsyncGRPCServer::ShutdownQueue() {
std::unique_lock<std::mutex> lock(cq_mutex_);
cq_send_->Shutdown();
cq_get_->Shutdown();
cq_prefetch_->Shutdown();
}
// This URL explains why shutdown is complicate:
......@@ -233,6 +239,7 @@ void AsyncGRPCServer::ShutDown() {
void AsyncGRPCServer::TryToRegisterNewSendOne() {
std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewSendOne";
return;
}
RequestSend* send = new RequestSend(&service_, cq_send_.get(), scope_,
......@@ -243,6 +250,7 @@ void AsyncGRPCServer::TryToRegisterNewSendOne() {
void AsyncGRPCServer::TryToRegisterNewGetOne() {
std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewGetOne";
return;
}
RequestGet* get = new RequestGet(&service_, cq_get_.get(), scope_, dev_ctx_,
......@@ -253,6 +261,7 @@ void AsyncGRPCServer::TryToRegisterNewGetOne() {
void AsyncGRPCServer::TryToRegisterNewPrefetchOne() {
std::unique_lock<std::mutex> lock(cq_mutex_);
if (is_shut_down_) {
VLOG(3) << "shutdown, do not TryToRegisterNewPrefetchOne";
return;
}
RequestPrefetch* prefetch =
......@@ -270,25 +279,28 @@ void AsyncGRPCServer::HandleRequest(::grpc::ServerCompletionQueue* cq,
void* tag = NULL;
bool ok = false;
while (true) {
VLOG(3) << "HandleRequest for " << cq_name << " while in";
if (!cq->Next(&tag, &ok)) {
LOG(INFO) << cq_name << " CompletionQueue shutdown!";
break;
}
VLOG(3) << "HandleRequest for " << cq_name << " while after Next";
PADDLE_ENFORCE(tag);
// 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_send") WaitCond(0);
RequestBase* base = (RequestBase*)tag;
RequestBase* base = reinterpret_cast<RequestBase*>(tag);
// reference:
// 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/ywATt88Ef_I
if (!ok) {
LOG(WARNING) << cq_name << " recv no regular event:argument name"
<< base->GetReqName();
LOG(WARNING) << cq_name << " recv no regular event:argument name["
<< base->GetReqName() << "]";
TryToRegisterNewOne();
delete base;
continue;
......
paddle/fluid/operators/detail/grpc_server.h
浏览文件 @ 19db989e
......@@ -15,7 +15,8 @@ limitations under the License. */
#pragma once
#include <grpc++/grpc++.h>
#include <thread>
#include <string>
#include <utility>
#include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/lod_tensor.h"
......@@ -93,6 +94,7 @@ class AsyncGRPCServer final {
// received variable from RPC, operators fetch variable from this queue.
SimpleBlockQueue<MessageWithName> var_get_queue_;
// client send variable to this queue.
ReceivedQueue var_recv_queue_;
// condition of the sub program
......
paddle/fluid/operators/detail/grpc_server_test.cc
浏览文件 @ 19db989e
......@@ -28,6 +28,7 @@ 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) {
......@@ -39,13 +40,23 @@ TEST(PREFETCH, CPU) {
platform::CPUPlace place;
platform::CPUDeviceContext ctx(place);
// create var on local scope
std::string var_name("tmp_0");
auto var = scope.Var(var_name);
auto tensor = var->GetMutable<framework::LoDTensor>();
tensor->Resize({10, 10});
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, var_name, "");
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 @ 19db989e
......@@ -80,7 +80,7 @@ enum class GrpcMethod {
};
static const int kGrpcNumMethods =
static_cast<int>(GrpcMethod::kGetVariable) + 1;
static_cast<int>(GrpcMethod::kPrefetchVariable) + 1;
inline const char* GrpcMethodName(GrpcMethod id) {
switch (id) {
......@@ -89,7 +89,7 @@ inline const char* GrpcMethodName(GrpcMethod id) {
case GrpcMethod::kGetVariable:
return "/sendrecv.SendRecvService/GetVariable";
case GrpcMethod::kPrefetchVariable:
return "/sendrecv.SendREcvService/PrefetchVariable";
return "/sendrecv.SendRecvService/PrefetchVariable";
}
// Shouldn't be reached.
......@@ -117,5 +117,5 @@ class GrpcService final {
};
} // namespace detail
} // namespace operator
} // namespace operators
} // namespace paddle
paddle/fluid/operators/detail/variable_response.cc
浏览文件 @ 19db989e
......@@ -13,7 +13,11 @@
// limitations under the License.
#include "paddle/fluid/operators/detail/variable_response.h"
#include <string.h>
#include <string>
#include <utility>
#include <vector>
#include "paddle/fluid/operators/detail/send_recv.pb.h"
#include "paddle/fluid/operators/detail/sendrecvop_utils.h"
......@@ -151,7 +155,7 @@ bool VariableResponse::CopySelectRowsTensorData(
auto* tensor = slr->mutable_value();
tensor->Resize(dims);
PADDLE_ENFORCE_EQ(
tensor->numel(),
static_cast<size_t>(tensor->numel()),
length / framework::SizeOfType(
paddle::operators::detail::ToTypeIndex(meta_.data_type())));
void* tensor_data = tensor->mutable_data(
......
paddle/fluid/operators/listen_and_serv_op.cc
浏览文件 @ 19db989e
......@@ -13,22 +13,13 @@ See the License for the specific language governing permissions and
limitations under the License. */
#include <stdint.h>
#include <sys/stat.h>
#include <ostream>
#include <thread>
#include <unistd.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/op_registry.h"
#include "paddle/fluid/framework/proto_desc.h"
#include "paddle/fluid/framework/threadpool.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 operators {
......@@ -54,16 +45,19 @@ static void CreateTensorFromMessageType(framework::Variable *var,
}
}
static void ParallelExecuteBlocks(const std::vector<size_t> &parallel_blkids,
framework::Executor *executor,
framework::ProgramDesc *program,
framework::Scope *scope) {
static void ParallelExecuteBlocks(
const std::vector<size_t> &parallel_blkids, framework::Executor *executor,
const std::vector<std::shared_ptr<framework::ExecutorPrepareContext>>
&prepared,
framework::ProgramDesc *program, framework::Scope *scope) {
std::vector<std::future<void>> fs;
for (size_t idx : parallel_blkids) {
fs.push_back(framework::Async([&executor, &program, &scope, idx]() {
fs.push_back(
framework::Async([&executor, &prepared, &program, &scope, idx]() {
int run_block = idx; // thread local
try {
executor->Run(*program, scope, run_block, false, false);
executor->RunPreparedContext(prepared[run_block].get(), scope,
false, false);
} catch (std::exception &e) {
LOG(ERROR) << "run sub program error " << e.what();
}
......@@ -105,11 +99,23 @@ class ListenAndServOp : public framework::OperatorBase {
auto *block = Attr<framework::BlockDesc *>(kOptimizeBlock);
auto *program = block->Program();
int num_blocks = program->Size();
size_t num_blocks = program->Size();
PADDLE_ENFORCE_GE(num_blocks, 2,
"server program should have at least 2 blocks");
framework::Executor executor(dev_place);
std::vector<int> block_list;
for (size_t blkid = 1; blkid < num_blocks; ++blkid)
block_list.push_back(blkid);
auto prepared = executor.Prepare(*program, block_list);
prepared.insert(
prepared.begin(),
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.
bool exit_flag = false;
......@@ -157,23 +163,25 @@ class ListenAndServOp : public framework::OperatorBase {
// The optimize blocks which have the same parent ID would run parallel
// TODO(Yancey1989): need to use ParallelExecutor for future
size_t last_parent_blkid = program->Block(1).Parent();
int32_t last_parent_blkid = program->Block(1).Parent();
std::vector<size_t> parallel_blkids;
parallel_blkids.push_back(1);
double ts = detail::GetTimestamp();
for (size_t blkid = 2; blkid < num_blocks; ++blkid) {
if (program->Block(blkid).Parent() != last_parent_blkid) {
for (size_t idx : parallel_blkids) VLOG(3) << idx;
ParallelExecuteBlocks(parallel_blkids, &executor, program,
ParallelExecuteBlocks(parallel_blkids, &executor, prepared, program,
&recv_scope);
parallel_blkids.clear();
last_parent_blkid = program->Block(blkid).Parent();
}
parallel_blkids.push_back(blkid);
}
ParallelExecuteBlocks(parallel_blkids, &executor, program, &recv_scope);
ParallelExecuteBlocks(parallel_blkids, &executor, prepared, program,
&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
// sum the input sparse variables which rows is empty at the next
......@@ -184,7 +192,8 @@ class ListenAndServOp : public framework::OperatorBase {
var->GetMutable<framework::SelectedRows>()->mutable_rows()->clear();
}
rpc_service_->SetCond(1);
// FIXME(typhoonzero): use another condition to sync wait clients get.
// NOTE: does not consider barrier request retry in here, we may use
// global barrier id to resolve this.
rpc_service_->WaitClientGet(fan_in);
sparse_vars.clear();
} // while(true)
......
paddle/fluid/operators/lookup_table_op.cc
浏览文件 @ 19db989e
......@@ -18,6 +18,22 @@ limitations under the License. */
namespace paddle {
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 {
public:
using framework::OperatorWithKernel::OperatorWithKernel;
......@@ -51,9 +67,7 @@ class LookupTableOp : public framework::OperatorWithKernel {
protected:
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext& ctx) const override {
return framework::OpKernelType(
framework::ToDataType(ctx.Input<LoDTensor>("W")->type()),
ctx.device_context());
return ExpectedKernelType(ctx);
}
};
......@@ -84,7 +98,7 @@ class LookupTableOpMaker : public framework::OpProtoAndCheckerMaker {
"If the value is -1, it makes no effect to lookup. "
"Otherwise the given value indicates padding the output "
"with zeros whenever lookup encounters it in Ids.")
.SetDefault(-1);
.SetDefault(kNoPadding);
AddComment(R"DOC(
Lookup Table Operator.
......@@ -124,9 +138,7 @@ class LookupTableOpGrad : public framework::OperatorWithKernel {
protected:
framework::OpKernelType GetExpectedKernelType(
const framework::ExecutionContext& ctx) const override {
return framework::OpKernelType(
framework::ToDataType(ctx.Input<LoDTensor>("W")->type()),
ctx.device_context());
return ExpectedKernelType(ctx);
}
};
......
paddle/fluid/operators/lookup_table_op.h
浏览文件 @ 19db989e
......@@ -14,6 +14,9 @@ limitations under the License. */
#pragma once
#include <string>
#include <vector>
#include "paddle/fluid/framework/eigen.h"
#include "paddle/fluid/framework/lod_tensor.h"
#include "paddle/fluid/framework/op_registry.h"
......@@ -25,16 +28,37 @@ namespace operators {
using Tensor = framework::Tensor;
using LoDTensor = framework::LoDTensor;
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>
class LookupTableKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& context) const override {
auto* table_t = context.Input<LoDTensor>("W");
auto* ids_var = context.InputVar("Ids");
Tensor* output_t = context.Output<Tensor>("Out");
void Compute(const framework::ExecutionContext &context) const override {
auto *table_var = context.InputVar("W");
auto *ids_var = context.InputVar("Ids");
Tensor *output_t = context.Output<Tensor>("Out");
int64_t padding_idx = context.Attr<int64_t>("padding_idx");
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");
}
int64_t* ids;
int64_t *ids;
int64_t ids_numel;
// The type of Ids(Input) is SelectedRows or LoDTensor, when Ids's type
......@@ -42,39 +66,50 @@ class LookupTableKernel : public framework::OpKernel<T> {
// when Ids's type is SelectedRows, the rows of Ids contains the
// ids to be looked up in W.
if (ids_var->IsType<LoDTensor>()) {
auto* ids_t = context.Input<LoDTensor>("Ids");
ids = const_cast<int64_t*>(ids_t->data<int64_t>());
auto *ids_t = context.Input<LoDTensor>("Ids");
ids = const_cast<int64_t *>(ids_t->data<int64_t>());
ids_numel = ids_t->numel();
} else if (ids_var->IsType<SelectedRows>()) {
auto* ids_t = context.Input<SelectedRows>("Ids");
ids = const_cast<int64_t*>(ids_t->rows().data());
auto *ids_t = context.Input<SelectedRows>("Ids");
ids = const_cast<int64_t *>(ids_t->rows().data());
ids_numel = ids_t->rows().size();
output_t->Resize({ids_numel, table_t->dims()[1]});
output_t->Resize({ids_numel, table_dim[1]});
} else {
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];
int D = table_t->dims()[1];
auto* table = table_t->data<T>();
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) {
PADDLE_ENFORCE_LT(ids[i], N);
if (padding_idx != kNoPadding && ids[i] == padding_idx) {
memset(output + i * row_width, 0, row_width * sizeof(T));
} else {
PADDLE_ENFORCE_LT(ids[i], row_number);
PADDLE_ENFORCE_GE(ids[i], 0);
memcpy(output + i * D, table + ids[i] * D, D * sizeof(T));
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) {
if (ids[i] == padding_idx) {
memset(output + i * D, 0, D * sizeof(T));
if (padding_idx != kNoPadding && ids[i] == padding_idx) {
memset(output + i * row_width, 0, row_width * sizeof(T));
} else {
PADDLE_ENFORCE_LT(ids[i], N);
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> {
template <typename T>
class LookupTableGradKernel : public framework::OpKernel<T> {
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");
// 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.
if (is_sparse) {
auto* ids = context.Input<LoDTensor>("Ids");
auto* table = context.Input<LoDTensor>("W");
auto* d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
auto* d_table = context.Output<SelectedRows>(framework::GradVarName("W"));
auto *ids = context.Input<LoDTensor>("Ids");
auto *d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
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();
framework::Vector<int64_t> new_rows;
......@@ -104,31 +149,30 @@ class LookupTableGradKernel : public framework::OpKernel<T> {
}
d_table->set_rows(new_rows);
auto* d_table_value = d_table->mutable_value();
d_table_value->Resize({ids_dim[0], table->dims()[1]});
auto *d_table_value = d_table->mutable_value();
d_table_value->Resize({ids_dim[0], table_dim[1]});
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_table_data = d_table_value->data<T>();
auto *d_output_data = d_output->data<T>();
auto *d_table_data = d_table_value->data<T>();
PADDLE_ENFORCE_EQ(d_table_value->dims(), d_output->dims());
memcpy(d_table_data, d_output_data, sizeof(T) * d_output->numel());
} else {
auto* ids = context.Input<LoDTensor>("Ids");
auto* d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
auto* d_table = context.Output<LoDTensor>(framework::GradVarName("W"));
auto* table = context.Input<LoDTensor>("W");
auto *ids = context.Input<LoDTensor>("Ids");
auto *d_output = context.Input<LoDTensor>(framework::GradVarName("Out"));
auto *d_table = context.Output<LoDTensor>(framework::GradVarName("W"));
auto* ids_data = ids->data<int64_t>();
auto *ids_data = ids->data<int64_t>();
auto ids_dim = ids->dims();
int N = table->dims()[0];
int N = table_dim[0];
int D = d_output->dims()[1];
auto* d_output_data = d_output->data<T>();
auto* d_table_data = d_table->mutable_data<T>(context.GetPlace());
auto *d_output_data = d_output->data<T>();
auto *d_table_data = d_table->mutable_data<T>(context.GetPlace());
memset(d_table_data, 0, d_table->numel() * sizeof(T));
......
paddle/fluid/operators/nccl_op_test.cu.cc
浏览文件 @ 19db989e
......@@ -137,6 +137,8 @@ class NCCLTester : public ::testing::Test {
TEST_F(NCCLTester, ncclInitOp) {}
// ncclAllReduceOp with desc
// TODO(helin): https://github.com/PaddlePaddle/Paddle/issues/9367
/*
TEST_F(NCCLTester, ncclAllReduceOp) {
std::unique_ptr<f::OpDesc> op2(new f::OpDesc);
op2->SetType("ncclAllReduce");
......@@ -184,6 +186,7 @@ TEST_F(NCCLTester, ncclAllReduceOp) {
}
}
}
*/
// ncclReduceOp with desc
TEST_F(NCCLTester, ncclReduceOp) {
......@@ -236,6 +239,8 @@ TEST_F(NCCLTester, ncclReduceOp) {
}
// ncclBcastOp with desc
// TODO(helin): https://github.com/PaddlePaddle/Paddle/issues/9540
/*
TEST_F(NCCLTester, ncclBcastOp) {
std::unique_ptr<f::OpDesc> op2(new f::OpDesc);
const int kRoot = 0;
......@@ -281,3 +286,4 @@ TEST_F(NCCLTester, ncclBcastOp) {
ASSERT_NEAR(ct[j], result, 1e-5);
}
}
*/
paddle/fluid/operators/reader/create_double_buffer_reader_op.cc
浏览文件 @ 19db989e
......@@ -20,12 +20,29 @@ namespace paddle {
namespace operators {
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 {
public:
struct Item {
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_;
platform::DeviceContext* ctx_;
......@@ -34,42 +51,44 @@ class DoubleBufferReader : public framework::DecoratedReader {
explicit DoubleBufferReader(
ReaderBase* reader, platform::Place target_place = platform::CPUPlace())
: DecoratedReader(reader), place_(target_place) {
for (size_t i = 0; i < kDoubleBufferSize; ++i) {
if (platform::is_gpu_place(place_)) {
#ifdef PADDLE_WITH_CUDA
for (size_t i = 0; i < kCacheSize; ++i) {
if (platform::is_gpu_place(place_)) {
ctxs_.emplace_back(new platform::CUDADeviceContext(
boost::get<platform::CUDAPlace>(place_)));
#endif
}
}
start_thread();
}
void start_thread() {
buffer_ = framework::MakeChannel<Item>(kDoubleBufferSize);
prefetcher_ = std::thread([this] { PrefetchThreadFunc(); });
#endif
StartPrefetcher();
}
bool HasNext() const override;
void ReadNext(std::vector<framework::LoDTensor>* out) override;
void ReInit() override;
~DoubleBufferReader() {
buffer_->Close();
prefetcher_.join();
delete buffer_;
~DoubleBufferReader() { EndPrefetcher(); }
private:
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();
std::thread prefetcher_;
framework::Channel<Item>* buffer_;
framework::Channel<Item>* channel_;
platform::Place place_;
std::vector<std::unique_ptr<platform::DeviceContext>> ctxs_;
mutable Item local_buffer_;
};
class CreateDoubleBufferReaderOp : public framework::OperatorBase {
......@@ -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) {
if (!HasNext()) {
PADDLE_THROW("There is no next data!");
}
if (local_buffer_.payloads_.empty()) {
buffer_->Receive(&local_buffer_);
}
*out = local_buffer_.payloads_;
local_buffer_.payloads_.clear();
if (local_buffer_.ctx_) {
local_buffer_.ctx_->Wait();
Item batch;
channel_->Receive(&batch);
*out = batch.payloads_;
if (batch.ctx_) {
batch.ctx_->Wait();
}
}
void DoubleBufferReader::ReInit() {
reader_->ReInit();
buffer_->Close();
prefetcher_.join();
delete buffer_;
start_thread();
EndPrefetcher();
StartPrefetcher();
}
void DoubleBufferReader::PrefetchThreadFunc() {
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()) {
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_)) {
std::vector<framework::LoDTensor> gpu_batch;
auto& gpu_ctx = this->ctxs_[gpu_ctx_offset++];
gpu_ctx_offset %= this->ctxs_.size();
gpu_batch.resize(batch.payloads_.size());
for (size_t i = 0; i < batch.payloads_.size(); ++i) {
framework::TensorCopy(batch.payloads_[i], place_, *gpu_ctx,
&gpu_batch[i]);
gpu_batch[i].set_lod(batch.payloads_[i].lod());
}
batch.ctx_ = gpu_ctx.get();
std::swap(gpu_batch, batch.payloads_);
auto& gpu_batch = gpu_tensor_cache[cached_tensor_id];
auto* gpu_ctx = ctxs_[cached_tensor_id].get();
gpu_batch.resize(cpu_batch.size());
for (size_t i = 0; i < cpu_batch.size(); ++i) {
framework::TensorCopy(cpu_batch[i], place_, *gpu_ctx, &gpu_batch[i]);
gpu_batch[i].set_lod(cpu_batch[i].lod());
}
batch.payloads_ = gpu_batch;
batch.ctx_ = gpu_ctx;
} else {
// CPUPlace
batch.payloads_ = cpu_batch;
}
++cached_tensor_id;
cached_tensor_id %= kCacheSize;
try {
buffer_->Send(&batch);
channel_->Send(&batch);
} catch (paddle::platform::EnforceNotMet e) {
VLOG(5) << "WARNING: The double buffer channel has been closed. The "
"prefetch thread will terminate.";
break;
}
}
buffer_->Close();
channel_->Close();
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 operators
} // namespace paddle
......
paddle/fluid/operators/reshape_op.cc
浏览文件 @ 19db989e
......@@ -17,90 +17,66 @@ limitations under the License. */
namespace paddle {
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 {
public:
ReshapeOpMaker(OpProto *proto, OpAttrChecker *op_checker)
: OpProtoAndCheckerMaker(proto, op_checker) {
AddInput("X", "The input tensor of reshape operator.");
AddOutput("Out", "The output tensor of reshape operator.");
AddAttr<std::vector<int>>("shape",
"(vector<int>) "
"Target shape of reshape operator.");
AddInput("X", "(Tensor). The input tensor of reshape operator.");
AddInput("Shape",
"(Tensor<int32>, optional). If provided, reshape according to "
"this given shape. That is to say it has a higher priority than "
"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",
"Change the source tensor's shape without copy memory.")
.SetDefault(true);
"(default: false) Change the source tensor's shape without "
"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(
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:
1. Given a 3-D tensor Input(X) with a shape [2, 4, 6], and the target shape
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.
2. Given a 3-D tensor Input(X) with a shape [2, 4, 6], and the target shape
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.
An example:
Given a 2-D tensor X with 2 rows and 2 columns : [[1, 2], [3, 4]]
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).
and target shape = [1, 4], the reshape operator will transform
the tensor X into a 2-D tensor: [[1, 2, 3, 4]]
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");
}
};
......@@ -119,6 +95,14 @@ class ReshapeGradOp : public framework::OperatorWithKernel {
"Input(Out@GRAD) shouldn't be null.");
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
......
paddle/fluid/operators/reshape_op.h
浏览文件 @ 19db989e
......@@ -20,17 +20,129 @@ limitations under the License. */
namespace paddle {
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>
class ReshapeKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& ctx) const {
auto* out = ctx.Output<framework::Tensor>("Out");
auto* in = ctx.Input<framework::Tensor>("X");
void Compute(const framework::ExecutionContext &ctx) const {
auto *out = ctx.Output<framework::LoDTensor>("Out");
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");
auto out_dims = out->dims();
out->Resize(out_dims);
if (!inplace) {
out->mutable_data<T>(ctx.GetPlace());
framework::TensorCopy(*in, ctx.GetPlace(), ctx.device_context(), out);
// TensorCopy will resize to in_dims.
out->Resize(out_dims);
} else {
out->ShareDataWith(*in);
......@@ -42,9 +154,10 @@ class ReshapeKernel : public framework::OpKernel<T> {
template <typename DeviceContext, typename T>
class ReshapeGradKernel : public framework::OpKernel<T> {
public:
void Compute(const framework::ExecutionContext& ctx) const {
auto* d_out = ctx.Input<framework::Tensor>(framework::GradVarName("Out"));
auto* d_x = ctx.Output<framework::Tensor>(framework::GradVarName("X"));
void Compute(const framework::ExecutionContext &ctx) const {
auto *d_out = ctx.Input<framework::Tensor>(framework::GradVarName("Out"));
auto *d_x = ctx.Output<framework::Tensor>(framework::GradVarName("X"));
d_x->mutable_data<T>(ctx.GetPlace());
bool inplace = ctx.Attr<bool>("inplace");
......
paddle/fluid/operators/split_ids_op.h
浏览文件 @ 19db989e
......@@ -30,19 +30,16 @@ class SplitIdsOpKernel : public framework::OpKernel<T> {
PADDLE_THROW("SplitIds do not support GPU kernel");
}
const auto* ids_t = ctx.Input<framework::LoDTensor>("Ids");
auto& ids_dims = ids_t->dims();
auto& ids_dims = ctx.Input<framework::LoDTensor>("Ids")->dims();
const T* ids = ctx.Input<framework::LoDTensor>("Ids")->data<T>();
auto outs = ctx.MultiOutput<framework::LoDTensor>("Out");
const T* ids = ids_t->data<T>();
const size_t shard_num = outs.size();
std::vector<std::vector<T>> out_ids;
out_ids.resize(outs.size());
// split id by their shard_num.
for (size_t i = 0; i < ids_dims[0]; ++i) {
for (int i = 0; i < ids_dims[0]; ++i) {
T id = ids[i];
size_t shard_id = static_cast<size_t>(id) % shard_num;
out_ids[shard_id].push_back(id);
......
paddle/fluid/pybind/pybind.cc
浏览文件 @ 19db989e
......@@ -504,10 +504,10 @@ All parameter, weight, gradient are variables in Paddle.
const std::unordered_set<std::string> &params,
const ProgramDesc &startup_program,
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,
params, startup_program, main_program,
loss_var_name, scope);
loss_var_name, scope, allow_op_delay);
})
.def("run", &ParallelExecutor::Run);
......
paddle/scripts/docker/build.sh
浏览文件 @ 19db989e
......@@ -104,8 +104,10 @@ EOF
# make install should also be test when unittest
make install -j `nproc`
pip install /usr/local/opt/paddle/share/wheels/*.whl
if [[ ${WITH_FLUID_ONLY:-OFF} == "OFF" ]] ; then
paddle version
fi
fi
}
......@@ -183,6 +185,14 @@ EOF
NCCL_DEPS=""
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
ADD python/dist/*.whl /
# run paddle version to install python packages first
......@@ -192,7 +202,7 @@ EOF
pip install /*.whl; apt-get install -f -y && \
apt-get clean -y && \
rm -f /*.whl && \
paddle version && \
${PADDLE_VERSION} && \
ldconfig
${DOCKERFILE_CUDNN_DSO}
${DOCKERFILE_GPU_ENV}
......@@ -200,7 +210,7 @@ EOF
ADD go/cmd/pserver/pserver /usr/bin/
ADD go/cmd/master/master /usr/bin/
# default command shows the paddle version and exit
CMD ["paddle", "version"]
CMD [${CMD}]
EOF
}
......
python/CMakeLists.txt
浏览文件 @ 19db989e
......@@ -81,6 +81,7 @@ if (WITH_TESTING)
# enable v2 API unittest only when paddle swig api is compiled
add_subdirectory(paddle/v2/tests)
add_subdirectory(paddle/v2/plot/tests)
add_subdirectory(paddle/v2/reader/tests)
endif()
endif()
add_subdirectory(paddle/fluid/tests)
......
python/paddle/dataset/__init__.py
浏览文件 @ 19db989e
......@@ -37,7 +37,7 @@ __all__ = [
'cifar',
'movielens',
'conll05',
'sentiment'
'sentiment',
'uci_housing',
'wmt14',
'wmt16',
......
python/paddle/fluid/distribute_transpiler.py
浏览文件 @ 19db989e
......@@ -276,12 +276,15 @@ class DistributeTranspiler:
suff_idx = v.name.find(".trainer_")
if suff_idx >= 0:
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,
persistable=True,
type=v.type,
dtype=v.dtype,
shape=v.shape)
if self.trainers > 1:
for trainer_id in xrange(self.trainers):
var = pserver_program.global_block().create_var(
name="%s.trainer_%d" % (orig_var_name, trainer_id),
......@@ -290,6 +293,8 @@ class DistributeTranspiler:
dtype=v.dtype,
shape=v.shape)
recv_inputs.append(var)
else:
recv_inputs.append(single_trainer_var)
# step3
optimize_block = pserver_program.create_block(0)
......@@ -511,8 +516,11 @@ class DistributeTranspiler:
def _append_split_op(self, program, gradblocks):
# 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(
program, gradblocks, add_trainer_suffix=True)
program, gradblocks, add_trainer_suffix=add_suffix)
for varname, splited_vars in var_mapping.iteritems():
# variable that don't need to split have empty splited_vars
if len(splited_vars) <= 1:
......
python/paddle/fluid/executor.py
浏览文件 @ 19db989e
......@@ -48,8 +48,7 @@ def as_numpy(tensor):
assert isinstance(tensor, core.LoDTensor)
lod = tensor.lod()
if len(lod) > 0:
raise RuntimeError(
"Some of your featched tensors hold LoD information. \
raise RuntimeError("Some of your fetched tensors hold LoD information. \
They can not be completely cast to Python ndarray. \
Please set the parameter 'return_numpy' as 'False' to \
return LoDTensor itself directly.")
......@@ -180,59 +179,23 @@ def get_program_cache_key(feed, fetch_list):
class Executor(object):
def __init__(self, places):
if not isinstance(places, list) and not isinstance(places, tuple):
places = [places]
act_places = []
for each in places:
def __init__(self, place):
self.place = place
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
p.set_place(place)
self.executor = core.Executor(p)
self.program_caches = dict()
def aslodtensor(self, data):
def accumulate(data):
if not isinstance(data, list):
return 1
return sum([accumulate(sub) for sub in data])
def parselod(data):
seq_lens = [accumulate(seq) for seq in data]
cur_len = 0
lod = [cur_len]
for l in seq_lens:
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])
def as_lodtensor(self, data):
if isinstance(data, list):
raise RuntimeError("Some of your feed data hold LoD information. \
They can not be completely cast from a list of Python \
ndarray to LoDTensor. Please convert data to LoDTensor \
directly before feeding the data.\
")
# single tensor case
tensor = core.LoDTensor()
tensor.set(flattened_data, self.places[0])
tensor.set_lod(lod)
tensor.set(data, self.place)
return tensor
def _get_program_cache(self, program_cache_key):
......@@ -293,7 +256,7 @@ class Executor(object):
feed_target_name = op.desc.output('Out')[0]
cur_feed = feed[feed_target_name]
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')
core.set_feed_variable(scope, cur_feed, feed_var_name, idx)
else:
......
python/paddle/fluid/framework.py
浏览文件 @ 19db989e
......@@ -847,6 +847,11 @@ class Block(object):
if not self.has_var(var.name()):
self.create_var(name=var.name(), desc=var, type=var.type())
# sync variables removed from c++ end
for var in self.vars.keys():
if not self.desc.find_var(var):
self.vars.pop(var)
# sync operators from cpp
ops_in_cpp = []
for op_idx in range(0, self.desc.op_size()):
......@@ -881,6 +886,19 @@ class Block(object):
op = Operator(self, op_desc)
self.ops.append(op)
# sync ops removed from c++ end
if end_index != -1 and end_index < len(self.ops):
ops_in_cpp_index = 0
ops_in_python_index = 0
while ops_in_python_index < len(
self.ops) and ops_in_cpp_index < len(ops_in_cpp):
if self.ops[ops_in_python_index].desc != ops_in_cpp[
ops_in_cpp_index]:
del self.ops[ops_in_python_index]
else:
ops_in_cpp_index += 1
ops_in_python_index += 1
assert len(self.ops) == len(ops_in_cpp)
for index in range(len(self.ops)):
assert self.ops[index].desc == ops_in_cpp[index]
......
python/paddle/fluid/layers/detection.py
浏览文件 @ 19db989e
......@@ -19,7 +19,6 @@ from layer_function_generator import generate_layer_fn
from layer_function_generator import autodoc
from ..layer_helper import LayerHelper
import tensor
import ops
import nn
import math
......@@ -130,9 +129,9 @@ def detection_output(loc,
target_box=loc,
code_type='decode_center_size')
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 = ops.reshape(x=scores, shape=old_shape)
scores = nn.reshape(x=scores, shape=old_shape)
scores = nn.transpose(scores, perm=[0, 2, 1])
scores.stop_gradient = True
nmsed_outs = helper.create_tmp_variable(dtype=decoded_box.dtype)
......@@ -463,7 +462,7 @@ def ssd_loss(location,
num, num_prior, num_class = confidence.shape
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.
# 1.1 Compute IOU similarity between ground-truth boxes and prior boxes.
......@@ -474,7 +473,7 @@ def ssd_loss(location,
# 2. Compute confidence for mining hard examples
# 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
target_label, _ = target_assign(
gt_label, matched_indices, mismatch_value=background_label)
......@@ -487,7 +486,7 @@ def ssd_loss(location,
conf_loss = nn.softmax_with_cross_entropy(confidence, target_label)
# 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
neg_indices = helper.create_tmp_variable(dtype='int32')
dtype = matched_indices.dtype
......@@ -556,7 +555,7 @@ def ssd_loss(location,
# 5.3 Compute overall weighted 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.
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)
if normalize:
normalizer = nn.reduce_sum(target_loc_weight)
......@@ -709,7 +708,7 @@ def multi_box_head(inputs,
new_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
def _is_list_or_tuple_(data):
......@@ -803,7 +802,7 @@ def multi_box_head(inputs,
mbox_loc.shape[0],
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)
# get conf
......@@ -819,7 +818,7 @@ def multi_box_head(inputs,
conf_loc.shape[0], conf_loc.shape[1] * conf_loc.shape[2] *
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)
if len(box_results) == 1:
......
python/paddle/fluid/layers/nn.py
浏览文件 @ 19db989e
......@@ -73,8 +73,10 @@ __all__ = [
'smooth_l1',
'one_hot',
'autoincreased_step_counter',
'reshape',
'lod_reset',
'lrn',
'pad',
]
......@@ -3271,6 +3273,8 @@ def one_hot(input, depth):
The one-hot tensor or LodTensor, same as input.
Examples:
.. code-block:: python
X is a LoDTensor:
X.lod = [[0, 1, 4]]
X.shape = [4, 1]
......@@ -3325,6 +3329,102 @@ def autoincreased_step_counter(counter_name=None, begin=1, step=1):
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):
"""
LoD Reset Operator. Set LoD of **x** to a new one specified by **y** or
......@@ -3488,3 +3588,62 @@ def lrn(input, n=5, k=1.0, alpha=1e-4, beta=0.75, name=None):
"beta": beta})
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 @ 19db989e
......@@ -49,7 +49,6 @@ __activations__ = [
__all__ = [
'mean',
'mul',
'reshape',
'scale',
'sigmoid_cross_entropy_with_logits',
'elementwise_add',
......
python/paddle/fluid/parallel_executor.py
浏览文件 @ 19db989e
......@@ -21,7 +21,11 @@ __all__ = ['ParallelExecutor']
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 = []
if use_cuda:
for i in xrange(core.get_cuda_device_count()):
......@@ -35,7 +39,12 @@ class ParallelExecutor(object):
places.append(p)
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()
main = framework.default_main_program()
......@@ -52,7 +61,8 @@ class ParallelExecutor(object):
startup.desc,
main.desc,
loss_name,
scope)
scope,
allow_op_delay)
self.scope = scope
def run(self, fetch_list):
......
python/paddle/fluid/tests/unittests/CMakeLists.txt
浏览文件 @ 19db989e
......@@ -29,6 +29,7 @@ function(py_test_modules TARGET_NAME)
endfunction()
# 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_dyn_rnn)
list(REMOVE_ITEM TEST_OPS test_mul_op)
......@@ -64,6 +65,7 @@ else()
endif(WITH_FAST_BUNDLE_TEST)
# 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_train_dyn_rnn MODULES test_dyn_rnn)
py_test_modules(test_mul_op MODULES test_mul_op)
......
python/paddle/fluid/tests/unittests/op_test.py
浏览文件 @ 19db989e
......@@ -334,7 +334,7 @@ class OpTest(unittest.TestCase):
np.allclose(
actual_t, expect_t, atol=atol),
"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):
self.assertListEqual(actual.lod(), expect[1],
"Output (" + out_name +
......@@ -568,6 +568,6 @@ class OpTest(unittest.TestCase):
fetch_list = [g for p, g in param_grad_list]
executor = Executor(place)
return map(
np.array,
executor.run(prog, feed_dict, fetch_list, return_numpy=False))
return map(np.array,
executor.run(prog, feed_dict, fetch_list,
return_numpy=False))
python/paddle/fluid/tests/unittests/test_recv_op.py
浏览文件 @ 19db989e
......@@ -23,7 +23,7 @@ import time
class TestRecvOp(unittest.TestCase):
def test_send(self):
def no_test_send(self):
# Run init_serv in a thread
place = fluid.CPUPlace()
p = Process(target=self.init_serv, args=(place, ))
......
python/paddle/fluid/tests/unittests/test_target_assign_op.py 100755 → 100644
浏览文件 @ 19db989e
文件模式从 100755 更改为 100644
python/paddle/v2/__init__.py
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/__init__.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/cifar.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/common.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/conll05.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/flowers.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/imdb.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/imikolov.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/mnist.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/movielens.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/mq2007.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/sentiment.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/cifar_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/common_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/flowers_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/imdb_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/imikolov_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/mnist_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/mq2007_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/test_sentiment.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/voc2012_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/tests/wmt16_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/uci_housing.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/voc2012.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/wmt14.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/dataset/wmt16.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/image.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/minibatch.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/__init__.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/creator.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/decorator.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/CMakeLists.txt 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/__init__.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/creator_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/decorator_test.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/test_data_creator.txt 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/test_reader_recordio.dat 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/reader/tests/test_recordio_creator.dat 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/tests/cat.jpg 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/paddle/v2/tests/test_image.py 0 → 100644
浏览文件 @ 19db989e
此差异已折叠。
python/setup.py.in
浏览文件 @ 19db989e
此差异已折叠。
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