From 8eae3262f3f184dbdf8549ec43ab7f76a90210cd Mon Sep 17 00:00:00 2001 From: chenhaozhe Date: Tue, 19 Jan 2021 10:49:10 +0800 Subject: [PATCH] fix some description about bert and yolov3 --- .../official/cv/yolov3_darknet53/README.md | 87 ++++++++++--------- model_zoo/official/nlp/bert/README.md | 7 +- .../nlp/bert/src/bert_for_finetune.py | 10 +++ 3 files changed, 59 insertions(+), 45 deletions(-) diff --git a/model_zoo/official/cv/yolov3_darknet53/README.md b/model_zoo/official/cv/yolov3_darknet53/README.md index 9181e11607..b366a9398b 100644 --- a/model_zoo/official/cv/yolov3_darknet53/README.md +++ b/model_zoo/official/cv/yolov3_darknet53/README.md @@ -1,26 +1,27 @@ # Contents -- [YOLOv3-DarkNet53 Description](#yolov3-darknet53-description) -- [Model Architecture](#model-architecture) -- [Dataset](#dataset) -- [Environment Requirements](#environment-requirements) -- [Quick Start](#quick-start) -- [Script Description](#script-description) - - [Script and Sample Code](#script-and-sample-code) - - [Script Parameters](#script-parameters) - - [Training Process](#training-process) - - [Training](#training) - - [Distributed Training](#distributed-training) - - [Evaluation Process](#evaluation-process) - - [Evaluation](#evaluation) -- [Model Description](#model-description) - - [Performance](#performance) - - [Evaluation Performance](#evaluation-performance) - - [Inference Performance](#evaluation-performance) -- [Description of Random Situation](#description-of-random-situation) -- [ModelZoo Homepage](#modelzoo-homepage) - -# [YOLOv3-DarkNet53 Description](#contents) +- [Contents](#contents) + - [YOLOv3-DarkNet53 Description](#yolov3-darknet53-description) + - [Model Architecture](#model-architecture) + - [Dataset](#dataset) + - [Environment Requirements](#environment-requirements) + - [Quick Start](#quick-start) + - [Script Description](#script-description) + - [Script and Sample Code](#script-and-sample-code) + - [Script Parameters](#script-parameters) + - [Training Process](#training-process) + - [Training](#training) + - [Distributed Training](#distributed-training) + - [Evaluation Process](#evaluation-process) + - [Evaluation](#evaluation) + - [Model Description](#model-description) + - [Performance](#performance) + - [Evaluation Performance](#evaluation-performance) + - [Inference Performance](#inference-performance) + - [Description of Random Situation](#description-of-random-situation) + - [ModelZoo Homepage](#modelzoo-homepage) + +## [YOLOv3-DarkNet53 Description](#contents) You only look once (YOLO) is a state-of-the-art, real-time object detection system. YOLOv3 is extremely fast and accurate. @@ -32,11 +33,11 @@ YOLOv3 uses a few tricks to improve training and increase performance, including [Paper](https://pjreddie.com/media/files/papers/YOLOv3.pdf): YOLOv3: An Incremental Improvement. Joseph Redmon, Ali Farhadi, University of Washington -# [Model Architecture](#contents) +## [Model Architecture](#contents) YOLOv3 use DarkNet53 for performing feature extraction, which is a hybrid approach between the network used in YOLOv2, Darknet-19, and that newfangled residual network stuff. DarkNet53 uses successive 3 × 3 and 1 × 1 convolutional layers and has some shortcut connections as well and is significantly larger. It has 53 convolutional layers. -# [Dataset](#contents) +## [Dataset](#contents) Note that you can run the scripts based on the dataset mentioned in original paper or widely used in relevant domain/network architecture. In the following sections, we will introduce how to run the scripts using the related dataset below. @@ -47,19 +48,19 @@ Dataset used: [COCO2014](https://cocodataset.org/#download) - Val：6GM, 40,504 images - Annotations: 241M, Train/Val annotations - Data format：zip files - - Note：Data will be processed in yolo_dataset.py, and unzip files before uses it. + - Note：Data will be processed in yolo_dataset.py, and unzip files before uses it. -# [Environment Requirements](#contents) +## [Environment Requirements](#contents) - Hardware（Ascend/GPU） - Prepare hardware environment with Ascend or GPU processor. If you want to try Ascend , please send the [application form](https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/file/other/Ascend%20Model%20Zoo%E4%BD%93%E9%AA%8C%E8%B5%84%E6%BA%90%E7%94%B3%E8%AF%B7%E8%A1%A8.docx) to ascend@huawei.com. Once approved, you can get the resources. - Framework - [MindSpore](https://www.mindspore.cn/install/en) - For more information, please check the resources below： - - [MindSpore Tutorials](https://www.mindspore.cn/tutorial/training/en/master/index.html) - - [MindSpore Python API](https://www.mindspore.cn/doc/api_python/en/master/index.html) + - [MindSpore Tutorials](https://www.mindspore.cn/tutorial/training/en/master/index.html) + - [MindSpore Python API](https://www.mindspore.cn/doc/api_python/en/master/index.html) -# [Quick Start](#contents) +## [Quick Start](#contents) After installing MindSpore via the official website, you can start training and evaluation in as follows. If running on GPU, please add `--device_target=GPU` in the python command or use the "_gpu" shell script ("xxx_gpu.sh"). @@ -101,9 +102,9 @@ python eval.py \ sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt ``` -# [Script Description](#contents) +## [Script Description](#contents) -## [Script and Sample Code](#contents) +### [Script and Sample Code](#contents) ```contents . @@ -134,7 +135,7 @@ sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt └─train.py # train net ``` -## [Script Parameters](#contents) +### [Script Parameters](#contents) ```parameters Major parameters in train.py as follow. @@ -197,9 +198,9 @@ optional arguments: Resize rate for multi-scale training. Default: None ``` -## [Training Process](#contents) +### [Training Process](#contents) -### Training +#### Training ```command python train.py \ @@ -230,7 +231,7 @@ After training, you'll get some checkpoint files under the outputs folder by def The model checkpoint will be saved in outputs directory. -### Distributed Training +#### Distributed Training For Ascend device, distributed training example(8p) by shell script @@ -261,9 +262,9 @@ epoch[319], iter[102300], loss:31.952403, 496.02 imgs/sec, lr:2.409552052995423e ... ``` -## [Evaluation Process](#contents) +### [Evaluation Process](#contents) -### Evaluation +#### Evaluation Before running the command below. If running on GPU, please add `--device_target=GPU` in the python command or use the "_gpu" shell script ("xxx_gpu.sh"). @@ -278,6 +279,8 @@ sh run_eval.sh dataset/coco2014/ checkpoint/0-319_102400.ckpt The above python command will run in the background. You can view the results through the file "log.txt". The mAP of the test dataset will be as follows: +This the standard format from `pycocotools`, you can refer to [cocodataset](https://cocodataset.org/#detection-eval) for more detail. + ```eval log # log.txt =============coco eval reulst========= @@ -295,11 +298,11 @@ The above python command will run in the background. You can view the results th Average Recall (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.551 ``` -# [Model Description](#contents) +## [Model Description](#contents) -## [Performance](#contents) +### [Performance](#contents) -### Evaluation Performance +#### Evaluation Performance | Parameters | YOLO |YOLO | | -------------------------- | ----------------------------------------------------------- |------------------------------------------------------------ | @@ -319,7 +322,7 @@ The above python command will run in the background. You can view the results th | Checkpoint for Fine tuning | 474M (.ckpt file) | 474M (.ckpt file) | | Scripts | https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/yolov3_darknet53 | https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/yolov3_darknet53 | -### Inference Performance +#### Inference Performance | Parameters | YOLO |YOLO | | ------------------- | --------------------------- |------------------------------| @@ -333,10 +336,10 @@ The above python command will run in the background. You can view the results th | Accuracy | 8pcs: 31.1% | 8pcs: 29.7%~30.3% (shape=416)| | Model for inference | 474M (.ckpt file) | 474M (.ckpt file) | -# [Description of Random Situation](#contents) +## [Description of Random Situation](#contents) There are random seeds in distributed_sampler.py, transforms.py, yolo_dataset.py files. -# [ModelZoo Homepage](#contents) +## [ModelZoo Homepage](#contents) Please check the official [homepage](https://gitee.com/mindspore/mindspore/tree/master/model_zoo). diff --git a/model_zoo/official/nlp/bert/README.md b/model_zoo/official/nlp/bert/README.md index 1fb895933f..41ce5e40c0 100644 --- a/model_zoo/official/nlp/bert/README.md +++ b/model_zoo/official/nlp/bert/README.md @@ -12,8 +12,8 @@ - [Pre-Training](#pre-training) - [Fine-Tuning and Evaluation](#fine-tuning-and-evaluation) - [Options and Parameters](#options-and-parameters) - - [Options:](#options) - - [Parameters:](#parameters) + - [Options](#options) + - [Parameters](#parameters) - [Training Process](#training-process) - [Training](#training) - [Running on Ascend](#running-on-ascend) @@ -380,7 +380,8 @@ config for lossscale and etc. ```text Parameters for dataset and network (Pre-Training/Fine-Tuning/Evaluation): seq_length length of input sequence: N, default is 128 - vocab_size size of each embedding vector: N, must be consistant with the dataset you use. Default is 21136 + vocab_size size of each embedding vector: N, must be consistant with the dataset you use. Default is 21128. + Usually, we use 21128 for CN vocabs and 30522 for EN vocabs according to the origin paper. hidden_size size of bert encoder layers: N, default is 768 num_hidden_layers number of hidden layers: N, default is 12 num_attention_heads number of attention heads: N, default is 12 diff --git a/model_zoo/official/nlp/bert/src/bert_for_finetune.py b/model_zoo/official/nlp/bert/src/bert_for_finetune.py index 3ef32cbc17..d8b37da579 100644 --- a/model_zoo/official/nlp/bert/src/bert_for_finetune.py +++ b/model_zoo/official/nlp/bert/src/bert_for_finetune.py @@ -50,6 +50,16 @@ def _tensor_grad_overflow(grad): class BertFinetuneCell(nn.Cell): """ Especifically defined for finetuning where only four inputs tensor are needed. + + Append an optimizer to the training network after that the construct + function can be called to create the backward graph. + + Different from the builtin loss_scale wrapper cell, we apply grad_clip before the optimization. + + Args: + network (Cell): The training network. Note that loss function should have been added. + optimizer (Optimizer): Optimizer for updating the weights. + scale_update_cell (Cell): Cell to do the loss scale. Default: None. """ def __init__(self, network, optimizer, scale_update_cell=None):