add note (#5427)

* add note

* fix log

* fix hflip in train

* fix readme

* fix serving doc

* fix link

docs/tipc/serving/serving.md

@@ -35,12 +35,14 @@ Paddle Serving依托深度学习框架PaddlePaddle旨在帮助深度学习开发
 <a name="21---"></a>
 ### 2.1 准备测试数据和部署环境
 
-【基本流程】
+**【基本流程】**
 
 **（1）准备测试数据：** 从验证集或者测试集中抽出至少一张图像，用于后续推理过程验证。
 
 **（2）准备部署环境**
 
+docker是一个开源的应用容器引擎，可以让应用程序更加方便地被打包和移植。建议在docker中进行Serving服务化部署。
+
 首先准备docker环境，AIStudio环境已经安装了合适的docker。如果是非AIStudio环境，请[参考文档](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.3/doc/doc_ch/environment.md#2)中的 "1.3.2 Docker环境配置" 安装docker环境。
 
 然后安装Paddle Serving三个安装包，paddle-serving-server，paddle-serving-client 和 paddle-serving-app。
@@ -68,7 +70,7 @@ Paddle Serving Server更多不同运行环境的whl包下载地址，请参考
 ```
 python3 -m paddle_serving_client.convert --dirname {静态图模型路径} --model_filename {模型结构文件} --params_filename {模型参数文件} --serving_server {转换后的服务器端模型和配置文件存储路径} --serving_client {转换后的客户端模型和配置文件存储路径}
 ```
-上面命令中 "转换后的服务器端模型和配置文件" 将用于后续服务化部署。
+上面命令中 "转换后的服务器端模型和配置文件" 将用于后续服务化部署。其中`paddle_serving_client.convert`命令是`paddle_serving_client` whl包内置的转换函数，无需修改。
 
 【实战】
 

docs/tipc/serving/template/code/pipeline_http_client.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import numpy as np
 import requests
 import json
@@ -7,6 +21,15 @@ import os
 
 
 def cv2_to_base64(image):
+    """cv2_to_base64
+
+    Convert an numpy array to a base64 object.
+
+    Args:
+        image: Input array.
+
+    Returns: Base64 output of the input.
+    """
     return base64.b64encode(image).decode('utf8')
 
 

docs/tipc/serving/template/code/web_service.py

@@ -16,26 +16,68 @@ from paddle_serving_server.web_service import WebService, Op
 
 
 class TIPCExampleOp(Op):
-    """
-    ExampleOp for serving server, you can rename by yourself
+    """TIPCExampleOp
+    
+    ExampleOp for serving server. You can rename by yourself. 
     """
 
     def init_op(self):
-        """
-        initialize the class
+        """init_op
+
+        Initialize the class.
+
+        Args: None
+
+        Returns: None
         """
         pass
 
     def preprocess(self, input_dicts, data_id, log_id):
-        # preprocess for the inputs
+        """preprocess
+        
+        In preprocess stage, assembling data for process stage. users can 
+        override this function for model feed features.
+
+        Args:
+            input_dicts: input data to be preprocessed
+            data_id: inner unique id, increase auto
+            log_id: global unique id for RTT, 0 default
+
+        Return:
+            output_data: data for process stage
+            is_skip_process: skip process stage or not, False default
+            prod_errcode: None default, otherwise, product errores occured.
+                          It is handled in the same way as exception. 
+            prod_errinfo: "" default
+        """
         pass
 
     def postprocess(self, input_dicts, fetch_dict, data_id, log_id):
+        """postprocess
+
+        In postprocess stage, assemble data for next op or output.
+        Args:
+            input_data: data returned in preprocess stage, dict(for single predict) or list(for batch predict)
+            fetch_data: data returned in process stage, dict(for single predict) or list(for batch predict)
+            data_id: inner unique id, increase auto
+            log_id: logid, 0 default
+
+        Returns: 
+            fetch_dict: fetch result must be dict type.
+            prod_errcode: None default, otherwise, product errores occured.
+                          It is handled in the same way as exception.
+            prod_errinfo: "" default
+        """
         # postprocess for the service output
         pass
 
 
 class TIPCExampleService(WebService):
+    """TIPCExampleService
+    
+    Service class to define the Serving OP.
+    """
+
     def get_pipeline_response(self, read_op):
         tipc_example_op = TIPCExampleOp(
             name="tipc_example", input_ops=[read_op])

docs/tipc/train_infer_python/template/code/export_model.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import paddle
 import paddle.nn as nn
 import paddle.nn.functional as F
@@ -9,6 +23,16 @@ import numpy as np
 
 # parse args
 def get_args(add_help=True):
+    """get_args
+
+    Parse all args using argparse lib
+
+    Args:
+        add_help: Whether to add -h option on args
+
+    Returns:
+        An object which contains many parameters used for inference.
+    """
     import argparse
     parser = argparse.ArgumentParser(
         description='PaddlePaddle Args', add_help=add_help)
@@ -17,14 +41,29 @@ def get_args(add_help=True):
 
 
 def build_model(args):
-    """
-    build model
+    """build_model
+
+    Build your own model.
+
+    Args:
+        args: Parameters generated using argparser.
+
+    Returns:
+        A model whose type is nn.Layer
     """
     pass
 
 
 def export(args):
-    # build your own model
+    """export
+
+    export inference model using jit.save
+
+    Args:
+        args: Parameters generated using argparser.
+
+    Returns: None
+    """
     model = build_model(args)
 
     # decorate model with jit.save

docs/tipc/train_infer_python/template/code/infer.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
 import os
 import paddle
 from paddle import inference
 import numpy as np
 from PIL import Image
 
-from reprod_log import ReprodLogger
-from preprocess_ops import ResizeImage, CenterCropImage, NormalizeImage, ToCHW, Compose
-
 
 class InferenceEngine(object):
+    """InferenceEngine
+    
+    Inference engina class which contains preprocess, run, postprocess
+
+    """
+
     def __init__(self, args):
+        """
+        Args:
+            args: Parameters generated using argparser.
+
+        Returns: None
+        """
         super().__init__()
         pass
 
     def load_predictor(self, model_file_path, params_file_path):
-        """
+        """load_predictor
+
         initialize the inference engine
+
+        Args:
+            model_file_path: inference model path (*.pdmodel)
+            model_file_path: inference parmaeter path (*.pdiparams)
+        Returns: None
         """
         pass
 
-    def preprocess(self, img_path):
-        # preprocess for data
+    def preprocess(self, x):
+        """preprocess
+
+        Preprocess to the input.
+
+        Args:
+            x: Raw input, it can be an image path, a numpy array and so on.
+
+        Returns: Input data after preprocess.
+        """
         pass
 
     def postprocess(self, x):
-        # postprocess for the inference engine output
+        """postprocess
+
+        Postprocess to the inference engine output.
+
+        Args:
+            x: Inference engine output.
+
+        Returns: Output data after postprocess.
+        """
         pass
 
     def run(self, x):
-        # run using the infer
+        """run
+
+        Inference process using inference engine.
+
+        Args:
+            x: Input data after preprocess.
+
+        Returns: Inference engine output
+        """
         pass
 
 
@@ -46,6 +99,17 @@ def get_args(add_help=True):
 
 
 def infer_main(args):
+    """infer_main
+
+    Main inference function.
+
+    Args:
+        args: Parameters generated using argparser.
+
+    Returns:
+        class_id: Class index of the input.
+        prob: : Probability of the input.
+    """
     # init inference engine
     inference_engine = InferenceEngine(args)
 
@@ -86,4 +150,4 @@ def infer_main(args):
 
 if __name__ == "__main__":
     args = get_args()
-    infer_main(args)
\ No newline at end of file
+    infer_main(args)

tutorials/mobilenetv3_prod/Step6/README.md

@@ -29,6 +29,13 @@
 
 在此感谢[vision](https://github.com/pytorch/vision)，提高了MobileNetV3论文复现的效率。
 
+注意：在这里为了简化流程，仅关于`ImageNet标准训练过程`做训练对齐，具体地：
+* 训练总共120epoch，总的batch size是256*8=2048，学习率为0.8，下降策略为Piecewise Decay(30epoch下降10倍)
+* 训练预处理：RandomResizedCrop(size=224) + RandomFlip(p=0.5) + Normalize
+* 评估预处理：Resize(256) + CenterCrop(224) + Normalize
+
+这里`mobilenet_v3_small`的参考指标也是重新训练得到的。
+
 ## 2. 数据集和复现精度
 
 数据集为ImageNet，训练集包含1281167张图像，验证集包含50000张图像。
@@ -38,9 +45,7 @@
 
 | 模型      | top1/5 acc (参考精度) | top1/5 acc (复现精度) | 下载链接 |
 |:---------:|:------:|:----------:|:----------:|
-| Mo | 0.677/0.874   | 0.677/0.874   | [预训练模型](https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/mobilenet_v3_small_paddle_pretrained.pdparams) \|  [Inference模型](https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/mobilenet_v3_small_paddle_infer.tar) \| [日志(coming soon)]() |
-
-* 注：目前提供的预训练模型是从参考代码提供的权重转过来的，完整的训练结果和日志敬请期待！
+| Mo | -/-   | 0.601/0.826   | [预训练模型](https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/mobilenet_v3_small_pretrained.pdparams) \|  [Inference模型(coming soon!)]() \| [日志](https://paddle-model-ecology.bj.bcebos.com/model/mobilenetv3_reprod/train_mobilenet_v3_small.log) |
 
 
 ## 3. 准备环境与数据
@@ -95,7 +100,7 @@ tar -xf test_images/lite_data.tar
 
 ```bash
 export CUDA_VISIBLE_DEVICES=0
-python3.7 train.py --data-path=./ILSVRC2012 --lr=0.00125 --batch-size=32
+python3.7 train.py --data-path=./ILSVRC2012 --lr=0.1 --batch-size=256
 ```
 
 部分训练日志如下所示。
@@ -109,7 +114,7 @@ python3.7 train.py --data-path=./ILSVRC2012 --lr=0.00125 --batch-size=32
 
 ```bash
 export CUDA_VISIBLE_DEVICES=0,1,2,3
-python3.7 -m paddle.distributed.launch --gpus="0,1,2,3" train.py --data-path="./ILSVRC2012" --lr=0.01 --batch-size=64
+python3.7 -m paddle.distributed.launch --gpus="0,1,2,3" train.py --data-path="./ILSVRC2012" --lr=0.4 --batch-size=256
 ```
 
 更多配置参数可以参考[train.py](./train.py)的`get_args_parser`函数。
@@ -146,7 +151,7 @@ python tools/predict.py --pretrained=./mobilenet_v3_small_paddle_pretrained.pdpa
     <img src="./images/demo.jpg" width=300">
 </div>
 
-最终输出结果为`class_id: 8, prob: 0.9503437280654907`，表示预测的类别ID是`8`，置信度为`0.950`。
+最终输出结果为`class_id: 8, prob: 0.9091238975524902`，表示预测的类别ID是`8`，置信度为`0.909`。
 
 * 使用CPU预测
 

tutorials/mobilenetv3_prod/Step6/paddlevision/transforms/functional.py

@@ -397,3 +397,19 @@ def resized_crop(
     img = crop(img, top, left, height, width)
     img = resize(img, size, interpolation)
     return img
+
+
+def hflip(img):
+    """Horizontally flip the given image.
+    Args:
+        img (PIL Image or Tensor): Image to be flipped. If img
+            is a Tensor, it is expected to be in [..., H, W] format,
+            where ... means it can have an arbitrary number of leading
+            dimensions.
+    Returns:
+        PIL Image or Tensor:  Horizontally flipped image.
+    """
+    if not isinstance(img, paddle.Tensor):
+        return F_pil.hflip(img)
+
+    return F_t.hflip(img)

tutorials/mobilenetv3_prod/Step6/paddlevision/transforms/functional_tensor.py

@@ -268,3 +268,7 @@ def resize(img: Tensor,
         out_dtype=out_dtype)
 
     return img
+
+
+def hflip(img):
+    return img.flip(-1)

tutorials/mobilenetv3_prod/Step6/paddlevision/transforms/transforms.py

 import math
 import numbers
+import random
 import warnings
 from collections.abc import Sequence
 from typing import Tuple, List
@@ -17,7 +18,7 @@ from .functional import InterpolationMode, _interpolation_modes_from_int
 
 __all__ = [
     "Compose", "ToTensor", "Normalize", "Resize", "CenterCrop",
-    "RandomResizedCrop"
+    "RandomResizedCrop", "RandomHorizontalFlip"
 ]
 
 
@@ -370,3 +371,28 @@ def _setup_size(size, error_msg):
         raise ValueError(error_msg)
 
     return size
+
+
+class RandomHorizontalFlip(paddle.nn.Layer):
+    """Horizontally flip the given image randomly with a given probability.
+    If the image is torch Tensor, it is expected
+    to have [..., H, W] shape, where ... means an arbitrary number of leading
+    dimensions
+    Args:
+        p (float): probability of the image being flipped. Default value is 0.5
+    """
+
+    def __init__(self, p=0.5):
+        super().__init__()
+        self.p = p
+
+    def forward(self, img):
+        """
+        Args:
+            img (PIL Image or Tensor): Image to be flipped.
+        Returns:
+            PIL Image or Tensor: Randomly flipped image.
+        """
+        if random.random() < self.p:
+            return F.hflip(img)
+        return img

tutorials/mobilenetv3_prod/Step6/presets.py

@@ -10,8 +10,8 @@ class ClassificationPresetTrain:
                  auto_augment_policy=None,
                  random_erase_prob=0.0):
         trans = [transforms.RandomResizedCrop(crop_size)]
-        #if hflip_prob > 0:
-        #    trans.append(transforms.RandomHorizontalFlip(hflip_prob))
+        if hflip_prob > 0:
+            trans.append(transforms.RandomHorizontalFlip(hflip_prob))
         #if auto_augment_policy is not None:
         #    aa_policy = autoaugment.AutoAugmentPolicy(auto_augment_policy)
         #    trans.append(autoaugment.AutoAugment(policy=aa_policy))

tutorials/mobilenetv3_prod/Step6/shell/train.sh

+export CUDA_VISIBLE_DEVICES=0
+
+python3.7 train.py \
+    --data-path /paddle/data/ILSVRC2012/ \
+    --model mobilenet_v3_small \
+    --lr 0.1 \
+    --batch-size=256 \
+    --output-dir "./output/" \
+    --epochs 120 \
+    --workers=6

tutorials/mobilenetv3_prod/Step6/shell/train_dist.sh

@@ -5,7 +5,7 @@ python3.7 -m paddle.distributed.launch \
     train.py \
     --data-path /paddle/data/ILSVRC2012/ \
     --model mobilenet_v3_small \
-    --lr 0.4 \
+    --lr 0.8 \
     --batch-size=256 \
     --output-dir "./output/" \
     --epochs 120 \

tutorials/mobilenetv3_prod/Step6/train.py

@@ -221,7 +221,6 @@ def main(args):
         lr_scheduler.step()
         if paddle.distributed.get_rank() == 0:
             top1 = evaluate(model, criterion, data_loader_test, device=device)
-            best_top1 = max(best_top1, top1)
             if args.output_dir:
                 paddle.save(model.state_dict(),
                             os.path.join(args.output_dir,
@@ -233,6 +232,12 @@ def main(args):
                             os.path.join(args.output_dir, 'latest.pdparams'))
                 paddle.save(optimizer.state_dict(),
                             os.path.join(args.output_dir, 'latest.pdopt'))
+                if top1 > best_top1:
+                    best_top1 = top1
+                    paddle.save(model.state_dict(),
+                                os.path.join(args.output_dir, 'best.pdparams'))
+                    paddle.save(optimizer.state_dict(),
+                                os.path.join(args.output_dir, 'best.pdopt'))
 
     total_time = time.time() - start_time
     total_time_str = str(datetime.timedelta(seconds=int(total_time)))
@@ -286,7 +291,7 @@ def get_args_parser(add_help=True):
         type=float,
         help='decrease lr by a factor of lr-gamma')
     parser.add_argument(
-        '--print-freq', default=1, type=int, help='print frequency')
+        '--print-freq', default=10, type=int, help='print frequency')
     parser.add_argument('--output-dir', default='.', help='path where to save')
     parser.add_argument('--resume', default='', help='resume from checkpoint')
     parser.add_argument(