未验证 提交 10f7e519 编写于 作者: 文幕地方's avatar 文幕地方 提交者: GitHub

Merge pull request #6 from PaddlePaddle/develop

merge paddleocr
English | [简体中文](README_cn.md)
## Introduction
Many user hopes package the PaddleOCR service into an docker image, so that it can be quickly released and used in the docker or k8s environment.
This page provide some standardized code to achieve this goal. You can quickly publish the PaddleOCR project into a callable Restful API service through the following steps. (At present, the deployment based on the HubServing mode is implemented first, and author plans to increase the deployment of the PaddleServing mode in the futrue)
## 1. Prerequisites
You need to install the following basic components first:
a. Docker
b. Graphics driver and CUDA 10.0+(GPU)
c. NVIDIA Container Toolkit(GPU,Docker 19.03+ can skip this)
d. cuDNN 7.6+(GPU)
## 2. Build Image
a. Download PaddleOCR sourcecode
```
git clone https://github.com/PaddlePaddle/PaddleOCR.git
```
b. Goto Dockerfile directory(ps:Need to distinguish between cpu and gpu version, the following takes cpu as an example, gpu version needs to replace the keyword)
```
cd docker/cpu
```
c. Build image
```
docker build -t paddleocr:cpu .
```
## 3. Start container
a. CPU version
```
sudo docker run -dp 8866:8866 --name paddle_ocr paddleocr:cpu
```
b. GPU version (base on NVIDIA Container Toolkit)
```
sudo nvidia-docker run -dp 8866:8866 --name paddle_ocr paddleocr:gpu
```
c. GPU version (Docker 19.03++)
```
sudo docker run -dp 8866:8866 --gpus all --name paddle_ocr paddleocr:gpu
```
d. Check service status(If you can see the following statement then it means completed:Successfully installed ocr_system && Running on http://0.0.0.0:8866/)
```
docker logs -f paddle_ocr
```
## 4. Test
a. Calculate the Base64 encoding of the picture to be recognized (if you just test, you can use a free online tool, like:https://freeonlinetools24.com/base64-image/)
b. Post a service request(sample request in sample_request.txt)
```
curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"Input image Base64 encode(need to delete the code 'data:image/jpg;base64,')\"]}" http://localhost:8866/predict/ocr_system
```
c. Get resposne(If the call is successful, the following result will be returned)
```
{"msg":"","results":[[{"confidence":0.8403433561325073,"text":"约定","text_region":[[345,377],[641,390],[634,540],[339,528]]},{"confidence":0.8131805658340454,"text":"最终相遇","text_region":[[356,532],[624,530],[624,596],[356,598]]}]],"status":"0"}
```
[English](README.md) | 简体中文
## Docker化部署服务
在日常项目应用中,相信大家一般都会希望能通过Docker技术,把PaddleOCR服务打包成一个镜像,以便在Docker或k8s环境里,快速发布上线使用。
本文将提供一些标准化的代码来实现这样的目标。大家通过如下步骤可以把PaddleOCR项目快速发布成可调用的Restful API服务。(目前暂时先实现了基于HubServing模式的部署,后续作者计划增加PaddleServing模式的部署)
## 1.实施前提准备
需要先完成如下基本组件的安装:
a. Docker环境
b. 显卡驱动和CUDA 10.0+(GPU)
c. NVIDIA Container Toolkit(GPU,Docker 19.03以上版本可以跳过此步)
d. cuDNN 7.6+(GPU)
## 2.制作镜像
a.下载PaddleOCR项目代码
```
git clone https://github.com/PaddlePaddle/PaddleOCR.git
```
b.切换至Dockerfile目录(注:需要区分cpu或gpu版本,下文以cpu为例,gpu版本需要替换一下关键字即可)
```
cd docker/cpu
```
c.生成镜像
```
docker build -t paddleocr:cpu .
```
## 3.启动Docker容器
a. CPU 版本
```
sudo docker run -dp 8866:8866 --name paddle_ocr paddleocr:cpu
```
b. GPU 版本 (通过NVIDIA Container Toolkit)
```
sudo nvidia-docker run -dp 8866:8866 --name paddle_ocr paddleocr:gpu
```
c. GPU 版本 (Docker 19.03以上版本,可以直接用如下命令)
```
sudo docker run -dp 8866:8866 --gpus all --name paddle_ocr paddleocr:gpu
```
d. 检查服务运行情况(出现:Successfully installed ocr_system和Running on http://0.0.0.0:8866/等信息,表示运行成功)
```
docker logs -f paddle_ocr
```
## 4.测试服务
a. 计算待识别图片的Base64编码(如果只是测试一下效果,可以通过免费的在线工具实现,如:http://tool.chinaz.com/tools/imgtobase/)
b. 发送服务请求(可参见sample_request.txt中的值)
```
curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"填入图片Base64编码(需要删除'data:image/jpg;base64,')\"]}" http://localhost:8866/predict/ocr_system
```
c. 返回结果(如果调用成功,会返回如下结果)
```
{"msg":"","results":[[{"confidence":0.8403433561325073,"text":"约定","text_region":[[345,377],[641,390],[634,540],[339,528]]},{"confidence":0.8131805658340454,"text":"最终相遇","text_region":[[356,532],[624,530],[624,596],[356,598]]}]],"status":"0"}
```
# Docker化部署服务
在日常项目应用中,相信大家一般都会希望能通过Docker技术,把PaddleOCR服务打包成一个镜像,以便在Docker或k8s环境里,快速发布上线使用。
English | [简体中文](README_cn.md)
本文将提供一些标准化的代码来实现这样的目标。大家通过如下步骤可以把PaddleOCR项目快速发布成可调用的Restful API服务。(目前暂时先实现了基于HubServing模式的部署,后续作者计划增加PaddleServing模式的部署)
## Introduction
Many user hopes package the PaddleOCR service into an docker image, so that it can be quickly released and used in the docker or k8s environment.
## 1.实施前提准备
This page provide some standardized code to achieve this goal. You can quickly publish the PaddleOCR project into a callable Restful API service through the following steps. (At present, the deployment based on the HubServing mode is implemented first, and author plans to increase the deployment of the PaddleServing mode in the futrue)
需要先完成如下基本组件的安装:
a. Docker环境
b. 显卡驱动和CUDA 10.0+(GPU)
c. NVIDIA Container Toolkit(GPU,Docker 19.03以上版本可以跳过此步)
## 1. Prerequisites
You need to install the following basic components first:
a. Docker
b. Graphics driver and CUDA 10.0+(GPU)
c. NVIDIA Container Toolkit(GPU,Docker 19.03+ can skip this)
d. cuDNN 7.6+(GPU)
## 2.制作镜像
a.下载PaddleOCR项目代码
## 2. Build Image
a. Download PaddleOCR sourcecode
```
git clone https://github.com/PaddlePaddle/PaddleOCR.git
```
b.切换至Dockerfile目录(注:需要区分cpu或gpu版本,下文以cpu为例,gpu版本需要替换一下关键字即可
b. Goto Dockerfile directory(ps:Need to distinguish between cpu and gpu version, the following takes cpu as an example, gpu version needs to replace the keyword
```
cd docker/cpu
```
c.生成镜像
c. Build image
```
docker build -t paddleocr:cpu .
```
## 3.启动Docker容器
a. CPU 版本
## 3. Start container
a. CPU version
```
sudo docker run -dp 8866:8866 --name paddle_ocr paddleocr:cpu
```
b. GPU 版本 (通过NVIDIA Container Toolkit)
b. GPU version (base on NVIDIA Container Toolkit)
```
sudo nvidia-docker run -dp 8866:8866 --name paddle_ocr paddleocr:gpu
```
c. GPU 版本 (Docker 19.03以上版本,可以直接用如下命令)
c. GPU version (Docker 19.03++)
```
sudo docker run -dp 8866:8866 --gpus all --name paddle_ocr paddleocr:gpu
```
d. 检查服务运行情况(出现:Successfully installed ocr_system和Running on http://0.0.0.0:8866/等信息,表示运行成功
d. Check service status(If you can see the following statement then it means completed:Successfully installed ocr_system && Running on http://0.0.0.0:8866/
```
docker logs -f paddle_ocr
```
## 4.测试服务
a. 计算待识别图片的Base64编码(如果只是测试一下效果,可以通过免费的在线工具实现,如:http://tool.chinaz.com/tools/imgtobase/)
b. 发送服务请求(可参见sample_request.txt中的值)
## 4. Test
a. Calculate the Base64 encoding of the picture to be recognized (if you just test, you can use a free online tool, like:https://freeonlinetools24.com/base64-image/)
b. Post a service request(sample request in sample_request.txt)
```
curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"填入图片Base64编码(需要删除'data:image/jpg;base64,')\"]}" http://localhost:8866/predict/ocr_system
curl -H "Content-Type:application/json" -X POST --data "{\"images\": [\"Input image Base64 encode(need to delete the code 'data:image/jpg;base64,')\"]}" http://localhost:8866/predict/ocr_system
```
c. 返回结果(如果调用成功,会返回如下结果
c. Get resposne(If the call is successful, the following result will be returned
```
{"msg":"","results":[[{"confidence":0.8403433561325073,"text":"约定","text_region":[[345,377],[641,390],[634,540],[339,528]]},{"confidence":0.8131805658340454,"text":"最终相遇","text_region":[[356,532],[624,530],[624,596],[356,598]]}]],"status":"0"}
```
......@@ -129,6 +129,7 @@ def parse_args():
parser.add_argument("--det", type=str2bool, default=True)
parser.add_argument("--rec", type=str2bool, default=True)
parser.add_argument("--use_zero_copy_run", type=bool, default=False)
return parser.parse_args()
......
......@@ -257,6 +257,7 @@ class SimpleReader(object):
norm_img = process_image_srn(
img=img,
image_shape=self.image_shape,
char_ops=self.char_ops,
num_heads=self.num_heads,
max_text_length=self.max_text_length)
else:
......
......@@ -4,8 +4,10 @@ import numpy as np
import paddle.fluid as fluid
import paddle.fluid.layers as layers
# Set seed for CE
dropout_seed = None
encoder_data_input_fields = (
"src_word",
"src_pos",
"src_slf_attn_bias", )
def wrap_layer_with_block(layer, block_idx):
......@@ -45,25 +47,6 @@ def wrap_layer_with_block(layer, block_idx):
return layer_wrapper
def position_encoding_init(n_position, d_pos_vec):
"""
Generate the initial values for the sinusoid position encoding table.
"""
channels = d_pos_vec
position = np.arange(n_position)
num_timescales = channels // 2
log_timescale_increment = (np.log(float(1e4) / float(1)) /
(num_timescales - 1))
inv_timescales = np.exp(np.arange(
num_timescales)) * -log_timescale_increment
scaled_time = np.expand_dims(position, 1) * np.expand_dims(inv_timescales,
0)
signal = np.concatenate([np.sin(scaled_time), np.cos(scaled_time)], axis=1)
signal = np.pad(signal, [[0, 0], [0, np.mod(channels, 2)]], 'constant')
position_enc = signal
return position_enc.astype("float32")
def multi_head_attention(queries,
keys,
values,
......@@ -200,10 +183,7 @@ def multi_head_attention(queries,
weights = layers.softmax(product)
if dropout_rate:
weights = layers.dropout(
weights,
dropout_prob=dropout_rate,
seed=dropout_seed,
is_test=False)
weights, dropout_prob=dropout_rate, seed=None, is_test=False)
out = layers.matmul(weights, v)
return out
......@@ -235,7 +215,7 @@ def positionwise_feed_forward(x, d_inner_hid, d_hid, dropout_rate):
act="relu")
if dropout_rate:
hidden = layers.dropout(
hidden, dropout_prob=dropout_rate, seed=dropout_seed, is_test=False)
hidden, dropout_prob=dropout_rate, seed=None, is_test=False)
out = layers.fc(input=hidden, size=d_hid, num_flatten_dims=2)
return out
......@@ -259,10 +239,7 @@ def pre_post_process_layer(prev_out, out, process_cmd, dropout_rate=0.):
elif cmd == "d": # add dropout
if dropout_rate:
out = layers.dropout(
out,
dropout_prob=dropout_rate,
seed=dropout_seed,
is_test=False)
out, dropout_prob=dropout_rate, seed=None, is_test=False)
return out
......@@ -271,7 +248,7 @@ post_process_layer = pre_post_process_layer
def prepare_encoder(
src_word, #[b,t,c]
src_word, # [b,t,c]
src_pos,
src_vocab_size,
src_emb_dim,
......@@ -286,9 +263,8 @@ def prepare_encoder(
This module is used at the bottom of the encoder stacks.
"""
src_word_emb = src_word #layers.concat(res,axis=1)
src_word_emb = src_word
src_word_emb = layers.cast(src_word_emb, 'float32')
# print("src_word_emb",src_word_emb)
src_word_emb = layers.scale(x=src_word_emb, scale=src_emb_dim**0.5)
src_pos_enc = layers.embedding(
......@@ -299,7 +275,7 @@ def prepare_encoder(
src_pos_enc.stop_gradient = True
enc_input = src_word_emb + src_pos_enc
return layers.dropout(
enc_input, dropout_prob=dropout_rate, seed=dropout_seed,
enc_input, dropout_prob=dropout_rate, seed=None,
is_test=False) if dropout_rate else enc_input
......@@ -324,7 +300,7 @@ def prepare_decoder(src_word,
param_attr=fluid.ParamAttr(
name=word_emb_param_name,
initializer=fluid.initializer.Normal(0., src_emb_dim**-0.5)))
# print("target_word_emb",src_word_emb)
src_word_emb = layers.scale(x=src_word_emb, scale=src_emb_dim**0.5)
src_pos_enc = layers.embedding(
src_pos,
......@@ -334,16 +310,10 @@ def prepare_decoder(src_word,
src_pos_enc.stop_gradient = True
enc_input = src_word_emb + src_pos_enc
return layers.dropout(
enc_input, dropout_prob=dropout_rate, seed=dropout_seed,
enc_input, dropout_prob=dropout_rate, seed=None,
is_test=False) if dropout_rate else enc_input
# prepare_encoder = partial(
# prepare_encoder_decoder, pos_enc_param_name=pos_enc_param_names[0])
# prepare_decoder = partial(
# prepare_encoder_decoder, pos_enc_param_name=pos_enc_param_names[1])
def encoder_layer(enc_input,
attn_bias,
n_head,
......@@ -412,234 +382,6 @@ def encoder(enc_input,
return enc_output
def decoder_layer(dec_input,
enc_output,
slf_attn_bias,
dec_enc_attn_bias,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
cache=None,
gather_idx=None):
""" The layer to be stacked in decoder part.
The structure of this module is similar to that in the encoder part except
a multi-head attention is added to implement encoder-decoder attention.
"""
slf_attn_output = multi_head_attention(
pre_process_layer(dec_input, preprocess_cmd, prepostprocess_dropout),
None,
None,
slf_attn_bias,
d_key,
d_value,
d_model,
n_head,
attention_dropout,
cache=cache,
gather_idx=gather_idx)
slf_attn_output = post_process_layer(
dec_input,
slf_attn_output,
postprocess_cmd,
prepostprocess_dropout, )
enc_attn_output = multi_head_attention(
pre_process_layer(slf_attn_output, preprocess_cmd,
prepostprocess_dropout),
enc_output,
enc_output,
dec_enc_attn_bias,
d_key,
d_value,
d_model,
n_head,
attention_dropout,
cache=cache,
gather_idx=gather_idx,
static_kv=True)
enc_attn_output = post_process_layer(
slf_attn_output,
enc_attn_output,
postprocess_cmd,
prepostprocess_dropout, )
ffd_output = positionwise_feed_forward(
pre_process_layer(enc_attn_output, preprocess_cmd,
prepostprocess_dropout),
d_inner_hid,
d_model,
relu_dropout, )
dec_output = post_process_layer(
enc_attn_output,
ffd_output,
postprocess_cmd,
prepostprocess_dropout, )
return dec_output
def decoder(dec_input,
enc_output,
dec_slf_attn_bias,
dec_enc_attn_bias,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
caches=None,
gather_idx=None):
"""
The decoder is composed of a stack of identical decoder_layer layers.
"""
for i in range(n_layer):
dec_output = decoder_layer(
dec_input,
enc_output,
dec_slf_attn_bias,
dec_enc_attn_bias,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
cache=None if caches is None else caches[i],
gather_idx=gather_idx)
dec_input = dec_output
dec_output = pre_process_layer(dec_output, preprocess_cmd,
prepostprocess_dropout)
return dec_output
def make_all_inputs(input_fields):
"""
Define the input data layers for the transformer model.
"""
inputs = []
for input_field in input_fields:
input_var = layers.data(
name=input_field,
shape=input_descs[input_field][0],
dtype=input_descs[input_field][1],
lod_level=input_descs[input_field][2]
if len(input_descs[input_field]) == 3 else 0,
append_batch_size=False)
inputs.append(input_var)
return inputs
def make_all_py_reader_inputs(input_fields, is_test=False):
reader = layers.py_reader(
capacity=20,
name="test_reader" if is_test else "train_reader",
shapes=[input_descs[input_field][0] for input_field in input_fields],
dtypes=[input_descs[input_field][1] for input_field in input_fields],
lod_levels=[
input_descs[input_field][2]
if len(input_descs[input_field]) == 3 else 0
for input_field in input_fields
])
return layers.read_file(reader), reader
def transformer(src_vocab_size,
trg_vocab_size,
max_length,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
label_smooth_eps,
bos_idx=0,
use_py_reader=False,
is_test=False):
if weight_sharing:
assert src_vocab_size == trg_vocab_size, (
"Vocabularies in source and target should be same for weight sharing."
)
data_input_names = encoder_data_input_fields + \
decoder_data_input_fields[:-1] + label_data_input_fields
if use_py_reader:
all_inputs, reader = make_all_py_reader_inputs(data_input_names,
is_test)
else:
all_inputs = make_all_inputs(data_input_names)
# print("all inputs",all_inputs)
enc_inputs_len = len(encoder_data_input_fields)
dec_inputs_len = len(decoder_data_input_fields[:-1])
enc_inputs = all_inputs[0:enc_inputs_len]
dec_inputs = all_inputs[enc_inputs_len:enc_inputs_len + dec_inputs_len]
label = all_inputs[-2]
weights = all_inputs[-1]
enc_output = wrap_encoder(
src_vocab_size, 64, n_layer, n_head, d_key, d_value, d_model,
d_inner_hid, prepostprocess_dropout, attention_dropout, relu_dropout,
preprocess_cmd, postprocess_cmd, weight_sharing, enc_inputs)
predict = wrap_decoder(
trg_vocab_size,
max_length,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
dec_inputs,
enc_output, )
# Padding index do not contribute to the total loss. The weights is used to
# cancel padding index in calculating the loss.
if label_smooth_eps:
label = layers.label_smooth(
label=layers.one_hot(
input=label, depth=trg_vocab_size),
epsilon=label_smooth_eps)
cost = layers.softmax_with_cross_entropy(
logits=predict,
label=label,
soft_label=True if label_smooth_eps else False)
weighted_cost = cost * weights
sum_cost = layers.reduce_sum(weighted_cost)
token_num = layers.reduce_sum(weights)
token_num.stop_gradient = True
avg_cost = sum_cost / token_num
return sum_cost, avg_cost, predict, token_num, reader if use_py_reader else None
def wrap_encoder_forFeature(src_vocab_size,
max_length,
n_layer,
......@@ -662,44 +404,8 @@ def wrap_encoder_forFeature(src_vocab_size,
img
"""
if enc_inputs is None:
# This is used to implement independent encoder program in inference.
conv_features, src_pos, src_slf_attn_bias = make_all_inputs(
encoder_data_input_fields)
else:
conv_features, src_pos, src_slf_attn_bias = enc_inputs #
b, t, c = conv_features.shape
#"""
# insert cnn
#"""
#import basemodel
# feat = basemodel.resnet_50(img)
# mycrnn = basemodel.CRNN()
# feat = mycrnn.ocr_convs(img,use_cudnn=TrainTaskConfig.use_gpu)
# b, c, w, h = feat.shape
# src_word = layers.reshape(feat, shape=[-1, c, w * h])
#myconv8 = basemodel.conv8()
#feat = myconv8.net(img )
#b , c, h, w = feat.shape#h=6
#print(feat)
#layers.Print(feat,message="conv_feat",summarize=10)
#feat =layers.conv2d(feat,c,filter_size =[4 , 1],act="relu")
#feat = layers.pool2d(feat,pool_stride=(3,1),pool_size=(3,1))
#src_word = layers.squeeze(feat,axes=[2]) #src_word [-1,c,ww]
#feat = layers.transpose(feat, [0,3,1,2])
#src_word = layers.reshape(feat,[-1,w, c*h])
#src_word = layers.im2sequence(
# input=feat,
# stride=[1, 1],
# filter_size=[feat.shape[2], 1])
#layers.Print(src_word,message="src_word",summarize=10)
# print('feat',feat)
#print("src_word",src_word)
enc_input = prepare_encoder(
conv_features,
......@@ -749,43 +455,9 @@ def wrap_encoder(src_vocab_size,
img, src_pos, src_slf_attn_bias = enc_inputs
img
"""
if enc_inputs is None:
# This is used to implement independent encoder program in inference.
src_word, src_pos, src_slf_attn_bias = make_all_inputs(
encoder_data_input_fields)
else:
src_word, src_pos, src_slf_attn_bias = enc_inputs #
#"""
# insert cnn
#"""
#import basemodel
# feat = basemodel.resnet_50(img)
# mycrnn = basemodel.CRNN()
# feat = mycrnn.ocr_convs(img,use_cudnn=TrainTaskConfig.use_gpu)
# b, c, w, h = feat.shape
# src_word = layers.reshape(feat, shape=[-1, c, w * h])
#myconv8 = basemodel.conv8()
#feat = myconv8.net(img )
#b , c, h, w = feat.shape#h=6
#print(feat)
#layers.Print(feat,message="conv_feat",summarize=10)
#feat =layers.conv2d(feat,c,filter_size =[4 , 1],act="relu")
#feat = layers.pool2d(feat,pool_stride=(3,1),pool_size=(3,1))
#src_word = layers.squeeze(feat,axes=[2]) #src_word [-1,c,ww]
#feat = layers.transpose(feat, [0,3,1,2])
#src_word = layers.reshape(feat,[-1,w, c*h])
#src_word = layers.im2sequence(
# input=feat,
# stride=[1, 1],
# filter_size=[feat.shape[2], 1])
#layers.Print(src_word,message="src_word",summarize=10)
src_word, src_pos, src_slf_attn_bias = enc_inputs #
# print('feat',feat)
#print("src_word",src_word)
enc_input = prepare_decoder(
src_word,
src_pos,
......@@ -811,248 +483,3 @@ def wrap_encoder(src_vocab_size,
preprocess_cmd,
postprocess_cmd, )
return enc_output
def wrap_decoder(trg_vocab_size,
max_length,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
dec_inputs=None,
enc_output=None,
caches=None,
gather_idx=None,
bos_idx=0):
"""
The wrapper assembles together all needed layers for the decoder.
"""
if dec_inputs is None:
# This is used to implement independent decoder program in inference.
trg_word, trg_pos, trg_slf_attn_bias, trg_src_attn_bias, enc_output = \
make_all_inputs(decoder_data_input_fields)
else:
trg_word, trg_pos, trg_slf_attn_bias, trg_src_attn_bias = dec_inputs
dec_input = prepare_decoder(
trg_word,
trg_pos,
trg_vocab_size,
d_model,
max_length,
prepostprocess_dropout,
bos_idx=bos_idx,
word_emb_param_name="src_word_emb_table"
if weight_sharing else "trg_word_emb_table")
dec_output = decoder(
dec_input,
enc_output,
trg_slf_attn_bias,
trg_src_attn_bias,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
caches=caches,
gather_idx=gather_idx)
return dec_output
# Reshape to 2D tensor to use GEMM instead of BatchedGEMM
dec_output = layers.reshape(
dec_output, shape=[-1, dec_output.shape[-1]], inplace=True)
if weight_sharing:
predict = layers.matmul(
x=dec_output,
y=fluid.default_main_program().global_block().var(
"trg_word_emb_table"),
transpose_y=True)
else:
predict = layers.fc(input=dec_output,
size=trg_vocab_size,
bias_attr=False)
if dec_inputs is None:
# Return probs for independent decoder program.
predict = layers.softmax(predict)
return predict
def fast_decode(src_vocab_size,
trg_vocab_size,
max_in_len,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
beam_size,
max_out_len,
bos_idx,
eos_idx,
use_py_reader=False):
"""
Use beam search to decode. Caches will be used to store states of history
steps which can make the decoding faster.
"""
data_input_names = encoder_data_input_fields + fast_decoder_data_input_fields
if use_py_reader:
all_inputs, reader = make_all_py_reader_inputs(data_input_names)
else:
all_inputs = make_all_inputs(data_input_names)
enc_inputs_len = len(encoder_data_input_fields)
dec_inputs_len = len(fast_decoder_data_input_fields)
enc_inputs = all_inputs[0:enc_inputs_len] #enc_inputs tensor
dec_inputs = all_inputs[enc_inputs_len:enc_inputs_len +
dec_inputs_len] #dec_inputs tensor
enc_output = wrap_encoder(
src_vocab_size,
64, ##to do !!!!!????
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
enc_inputs,
bos_idx=bos_idx)
start_tokens, init_scores, parent_idx, trg_src_attn_bias = dec_inputs
def beam_search():
max_len = layers.fill_constant(
shape=[1],
dtype=start_tokens.dtype,
value=max_out_len,
force_cpu=True)
step_idx = layers.fill_constant(
shape=[1], dtype=start_tokens.dtype, value=0, force_cpu=True)
cond = layers.less_than(x=step_idx, y=max_len) # default force_cpu=True
while_op = layers.While(cond)
# array states will be stored for each step.
ids = layers.array_write(
layers.reshape(start_tokens, (-1, 1)), step_idx)
scores = layers.array_write(init_scores, step_idx)
# cell states will be overwrited at each step.
# caches contains states of history steps in decoder self-attention
# and static encoder output projections in encoder-decoder attention
# to reduce redundant computation.
caches = [
{
"k": # for self attention
layers.fill_constant_batch_size_like(
input=start_tokens,
shape=[-1, n_head, 0, d_key],
dtype=enc_output.dtype,
value=0),
"v": # for self attention
layers.fill_constant_batch_size_like(
input=start_tokens,
shape=[-1, n_head, 0, d_value],
dtype=enc_output.dtype,
value=0),
"static_k": # for encoder-decoder attention
layers.create_tensor(dtype=enc_output.dtype),
"static_v": # for encoder-decoder attention
layers.create_tensor(dtype=enc_output.dtype)
} for i in range(n_layer)
]
with while_op.block():
pre_ids = layers.array_read(array=ids, i=step_idx)
# Since beam_search_op dosen't enforce pre_ids' shape, we can do
# inplace reshape here which actually change the shape of pre_ids.
pre_ids = layers.reshape(pre_ids, (-1, 1, 1), inplace=True)
pre_scores = layers.array_read(array=scores, i=step_idx)
# gather cell states corresponding to selected parent
pre_src_attn_bias = layers.gather(
trg_src_attn_bias, index=parent_idx)
pre_pos = layers.elementwise_mul(
x=layers.fill_constant_batch_size_like(
input=pre_src_attn_bias, # cann't use lod tensor here
value=1,
shape=[-1, 1, 1],
dtype=pre_ids.dtype),
y=step_idx,
axis=0)
logits = wrap_decoder(
trg_vocab_size,
max_in_len,
n_layer,
n_head,
d_key,
d_value,
d_model,
d_inner_hid,
prepostprocess_dropout,
attention_dropout,
relu_dropout,
preprocess_cmd,
postprocess_cmd,
weight_sharing,
dec_inputs=(pre_ids, pre_pos, None, pre_src_attn_bias),
enc_output=enc_output,
caches=caches,
gather_idx=parent_idx,
bos_idx=bos_idx)
# intra-beam topK
topk_scores, topk_indices = layers.topk(
input=layers.softmax(logits), k=beam_size)
accu_scores = layers.elementwise_add(
x=layers.log(topk_scores), y=pre_scores, axis=0)
# beam_search op uses lod to differentiate branches.
accu_scores = layers.lod_reset(accu_scores, pre_ids)
# topK reduction across beams, also contain special handle of
# end beams and end sentences(batch reduction)
selected_ids, selected_scores, gather_idx = layers.beam_search(
pre_ids=pre_ids,
pre_scores=pre_scores,
ids=topk_indices,
scores=accu_scores,
beam_size=beam_size,
end_id=eos_idx,
return_parent_idx=True)
layers.increment(x=step_idx, value=1.0, in_place=True)
# cell states(caches) have been updated in wrap_decoder,
# only need to update beam search states here.
layers.array_write(selected_ids, i=step_idx, array=ids)
layers.array_write(selected_scores, i=step_idx, array=scores)
layers.assign(gather_idx, parent_idx)
layers.assign(pre_src_attn_bias, trg_src_attn_bias)
length_cond = layers.less_than(x=step_idx, y=max_len)
finish_cond = layers.logical_not(layers.is_empty(x=selected_ids))
layers.logical_and(x=length_cond, y=finish_cond, out=cond)
finished_ids, finished_scores = layers.beam_search_decode(
ids, scores, beam_size=beam_size, end_id=eos_idx)
return finished_ids, finished_scores
finished_ids, finished_scores = beam_search()
return finished_ids, finished_scores, reader if use_py_reader else None
Markdown is supported
0% .
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