!7951 Modelzoo retinaface, improvement of learning rate descent function.

Merge pull request !7951 from zhanghuiyao/4p_retinaface

model_zoo/official/cv/retinaface_resnet50/README.md

@@ -80,7 +80,7 @@ After installing MindSpore via the official website and download the dataset, yo
   python train.py > train.log 2>&1 & 
   
   # run distributed training example
-  bash scripts/run_distribute_gpu_train.sh 3 0,1,2
+  bash scripts/run_distribute_gpu_train.sh 4 0,1,2,3
   
   # run evaluation example
   export CUDA_VISIBLE_DEVICES=0
@@ -109,7 +109,8 @@ After installing MindSpore via the official website and download the dataset, yo
         │   ├──config.py            // parameter configuration 
         │   ├──augmentation.py     // data augment method 
         │   ├──loss.py            // loss function 
-        │   ├──utils.py          // data preprocessing 
+        │   ├──utils.py          // data preprocessing
+        │   ├──lr_schedule.py   // learning rate schedule
         ├── data 
         │   ├──widerface                    // dataset data
         │   ├──resnet50_pretrain.ckpt      // resnet50 imagenet pretrain model
@@ -136,7 +137,7 @@ Parameters for both training and evaluation can be set in config.py
     'batch_size': 8,                                          # Batch size of train
     'num_workers': 8,                                         # Num worker of dataset load data
     'num_anchor': 29126,                                      # Num of anchor boxes, it depends on the image size
-    'ngpu': 3,                                                # Num gpu of train
+    'ngpu': 4,                                                # Num gpu of train
     'epoch': 100,                                             # Training epoch number
     'decay1': 70,                                             # Epoch number of the first weight attenuation
     'decay2': 90,                                             # Epoch number of the second weight attenuation
@@ -146,29 +147,31 @@ Parameters for both training and evaluation can be set in config.py
     'out_channel': 256,                                       # Output channel of DetectionHead
     'match_thresh': 0.35,                                     # Threshold for match box
     'optim': 'sgd',                                           # Optimizer type
-    'warmup_epoch': -1,                                       # Warmup size, -1 means no warm-up
-    'initial_lr': 0.001,                                      # Learning rate
+    'warmup_epoch': 5,                                        # Warmup size, 0 means no warm-up
+    'initial_lr': 0.01,                                       # Learning rate
     'network': 'resnet50',                                    # Backbone name
     'momentum': 0.9,                                          # Momentum for Optimizer
     'weight_decay': 5e-4,                                     # Weight decay for Optimizer
     'gamma': 0.1,                                             # Attenuation ratio of learning rate
     'ckpt_path': './checkpoint/',                             # Model save path
-    'save_checkpoint_steps': 1000,                            # Save checkpoint steps
+    'save_checkpoint_steps': 2000,                            # Save checkpoint steps
     'keep_checkpoint_max': 1,                                 # Number of reserved checkpoints
     'resume_net': None,                                       # Network for restart, default is None
     'training_dataset': '',                                   # Training dataset label path, like 'data/widerface/train/label.txt'
     'pretrain': True,                                         # whether training based on the pre-trained backbone
     'pretrain_path': './data/res50_pretrain.ckpt',            # Pre-trained backbone checkpoint path
-    # val
-    'val_model': './checkpoint/ckpt_0/RetinaFace-100_536.ckpt', # Validation model path
-    'val_dataset_folder': './data/widerface/val/',            # Validation dataset path
-    'val_origin_size': False,                                 # Is full size verification used
-    'val_confidence_threshold': 0.02,                         # Threshold for val confidence
-    'val_nms_threshold': 0.4,                                 # Threshold for val NMS
-    'val_iou_threshold': 0.5,                                 # Threshold for val IOU
-    'val_save_result': False,                                 # Whether save the resultss
-    'val_predict_save_folder': './widerface_result',          # Result save path 
-    'val_gt_dir': './data/ground_truth/',                      # Path of val set ground_truth
+    'seed': 1,                                                # setup train seed
+    'lr_type': 'dynamic_lr',
+  # val
+    'val_model': './checkpoint/ckpt_0/RetinaFace-100_536.ckpt',   # Validation model path
+    'val_dataset_folder': './data/widerface/val/',                # Validation dataset path
+    'val_origin_size': False,                                     # Is full size verification used
+    'val_confidence_threshold': 0.02,                             # Threshold for val confidence
+    'val_nms_threshold': 0.4,                                     # Threshold for val NMS
+    'val_iou_threshold': 0.5,                                     # Threshold for val IOU
+    'val_save_result': False,                                     # Whether save the resultss
+    'val_predict_save_folder': './widerface_result',              # Result save path 
+    'val_gt_dir': './data/ground_truth/',                         # Path of val set ground_truth
   ```
 
 
@@ -193,7 +196,7 @@ Parameters for both training and evaluation can be set in config.py
 - running on GPU
 
   ```
-  bash scripts/run_distribute_gpu_train.sh 3 0,1,2
+  bash scripts/run_distribute_gpu_train.sh 4 0,1,2,3
   ```
   
   The above shell script will run distribute training in the background. You can view the results through the file `train/train.log`.
@@ -207,7 +210,7 @@ Parameters for both training and evaluation can be set in config.py
 
 - evaluation on WIDERFACE dataset when running on GPU
 
-  Before running the command below, please check the checkpoint path used for evaluation. Please set the checkpoint path to be the absolute full path in src/config.py, e.g., "username/retinaface/checkpoint/ckpt_0/RetinaFace-100_536.ckpt".
+  Before running the command below, please check the checkpoint path used for evaluation. Please set the checkpoint path to be the absolute full path in src/config.py, e.g., "username/retinaface/checkpoint/ckpt_0/RetinaFace-100_402.ckpt".
   
   ```
   export CUDA_VISIBLE_DEVICES=0
@@ -218,7 +221,7 @@ Parameters for both training and evaluation can be set in config.py
   
   ```
   # grep "Val AP" eval.log
-  Easy   Val AP : 0.9413
+  Easy   Val AP : 0.9422
   Medium Val AP : 0.9325
   Hard   Val AP : 0.8900
   ```
@@ -233,7 +236,7 @@ Parameters for both training and evaluation can be set in config.py
   
   ```
   # grep "Val AP" eval.log
-  Easy   Val AP : 0.9413
+  Easy   Val AP : 0.9422
   Medium Val AP : 0.9325
   Hard   Val AP : 0.8900
   ```
@@ -253,14 +256,14 @@ Parameters for both training and evaluation can be set in config.py
 | uploaded Date              | 10/16/2020 (month/day/year)                                  |
 | MindSpore Version          | 1.0.0                                                        |
 | Dataset                    | WIDERFACE                                                    |
-| Training Parameters        | epoch=100, steps=536, batch_size=8, lr=0.001              |
+| Training Parameters        | epoch=100, steps=402, batch_size=8, lr=0.01                  |
 | Optimizer                  | SGD                                                          |
 | Loss Function              | MultiBoxLoss + Softmax Cross Entropy                         |
 | outputs                    | bounding box + confidence + landmark                         |
 | Loss                       | 1.200                                                        |
-| Speed                      | 3pcs: 566 ms/step                                            |
-| Total time                 | 3pcs: 8.43 hours                                              |
-| Parameters (M)             | 27.29M                                                            |
+| Speed                      | 4pcs: 560 ms/step                                            |
+| Total time                 | 4pcs: 6.4 hours                                              |
+| Parameters (M)             | 27.29M                                                       |
 | Checkpoint for Fine tuning | 336.3M (.ckpt file)                                          |
 | Scripts                    | [retinaface script](https://gitee.com/mindspore/mindspore/tree/master/model_zoo/official/cv/retinaface) |
 

model_zoo/official/cv/retinaface_resnet50/scripts/run_distribute_gpu_train.sh

@@ -17,7 +17,7 @@
 echo "=============================================================================================================="
 echo "Please run the script as: "
 echo "bash run_distribute_gpu_train.sh DEVICE_NUM CUDA_VISIBLE_DEVICES"
-echo "for example: bash run_distribute_gpu_train.sh 3 0,1,2"
+echo "for example: bash run_distribute_gpu_train.sh 4 0,1,2,3"
 echo "=============================================================================================================="
 
 RANK_SIZE=$1

model_zoo/official/cv/retinaface_resnet50/src/config.py

@@ -25,30 +25,34 @@ cfg_res50 = {
     'batch_size': 8,
     'num_workers': 8,
     'num_anchor': 29126,
-    'ngpu': 3,
-    'epoch': 100,
-    'decay1': 70,
-    'decay2': 90,
+    'ngpu': 4,
     'image_size': 840,
     'return_layers': {'layer2': 1, 'layer3': 2, 'layer4': 3},
     'in_channel': 256,
     'out_channel': 256,
     'match_thresh': 0.35,
-    'optim': 'sgd',
-    'warmup_epoch': -1,
-    'initial_lr': 0.001,
     'network': 'resnet50',
 
     # opt
+    'optim': 'sgd',
     'momentum': 0.9,
     'weight_decay': 5e-4,
 
+    # seed
+    'seed': 1,
+
     # lr
+    'epoch': 100,
+    'decay1': 70,
+    'decay2': 90,
+    'lr_type': 'dynamic_lr',
+    'initial_lr': 0.01,
+    'warmup_epoch': 5,
     'gamma': 0.1,
 
     # checkpoint
     'ckpt_path': './checkpoint/',
-    'save_checkpoint_steps': 1000,
+    'save_checkpoint_steps': 2000,
     'keep_checkpoint_max': 1,
     'resume_net': None,
 

model_zoo/official/cv/retinaface_resnet50/src/lr_schedule.py

@@ -0,0 +1,61 @@
+# Copyright 2020 Huawei Technologies Co., Ltd
+#
+# 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
+#
+# less 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.
+# ============================================================================
+"""learning rate schedule."""
+import math
+from .config import cfg_res50
+
+
+def _linear_warmup_learning_rate(current_step, warmup_steps, base_lr, init_lr):
+    lr_inc = (float(base_lr) - float(init_lr)) / float(warmup_steps)
+    learning_rate = float(init_lr) + lr_inc * current_step
+    return learning_rate
+
+
+def _a_cosine_learning_rate(current_step, base_lr, warmup_steps, decay_steps):
+    base = float(current_step - warmup_steps) / float(decay_steps)
+    learning_rate = (1 + math.cos(base * math.pi)) / 2 * base_lr
+    return learning_rate
+
+
+def _dynamic_lr(base_lr, total_steps, warmup_steps, warmup_ratio=1 / 3):
+    lr = []
+    for i in range(total_steps):
+        if i < warmup_steps:
+            lr.append(_linear_warmup_learning_rate(i, warmup_steps, base_lr, base_lr * warmup_ratio))
+        else:
+            lr.append(_a_cosine_learning_rate(i, base_lr, warmup_steps, total_steps))
+
+    return lr
+
+
+def adjust_learning_rate(initial_lr, gamma, stepvalues, steps_pre_epoch, total_epochs, warmup_epoch=5):
+    if cfg_res50['lr_type'] == 'dynamic_lr':
+        return _dynamic_lr(initial_lr, total_epochs * steps_pre_epoch, warmup_epoch * steps_pre_epoch,
+                           warmup_ratio=1 / 3)
+
+    lr_each_step = []
+    for epoch in range(1, total_epochs + 1):
+        for _ in range(steps_pre_epoch):
+            if epoch <= warmup_epoch:
+                lr = 0.1 * initial_lr * (1.5849 ** (epoch - 1))
+            else:
+                if stepvalues[0] <= epoch <= stepvalues[1]:
+                    lr = initial_lr * (gamma ** (1))
+                elif epoch > stepvalues[1]:
+                    lr = initial_lr * (gamma ** (2))
+                else:
+                    lr = initial_lr
+            lr_each_step.append(lr)
+    return lr_each_step

model_zoo/official/cv/retinaface_resnet50/train.py

@@ -14,12 +14,9 @@
 # ============================================================================
 """Train Retinaface_resnet50."""
 from __future__ import print_function
-import random
 import math
-import numpy as np
+import mindspore
 
-import mindspore.nn as nn
-import mindspore.dataset as de
 from mindspore import context
 from mindspore.context import ParallelMode
 from mindspore.train import Model
@@ -31,28 +28,7 @@ from src.config import cfg_res50
 from src.network import RetinaFace, RetinaFaceWithLossCell, TrainingWrapper, resnet50
 from src.loss import MultiBoxLoss
 from src.dataset import create_dataset
-
-def setup_seed(seed):
-    random.seed(seed)
-    np.random.seed(seed)
-    de.config.set_seed(seed)
-
-def adjust_learning_rate(initial_lr, gamma, stepvalues, steps_per_epoch, total_epochs, warmup_epoch=5):
-    lr_each_step = []
-    for epoch in range(1, total_epochs+1):
-        for step in range(steps_per_epoch):
-            if epoch <= warmup_epoch:
-                lr = 1e-6 + (initial_lr - 1e-6) * ((epoch - 1) * steps_per_epoch + step) / \
-                     (steps_per_epoch * warmup_epoch)
-            else:
-                if stepvalues[0] <= epoch <= stepvalues[1]:
-                    lr = initial_lr * (gamma ** (1))
-                elif epoch > stepvalues[1]:
-                    lr = initial_lr * (gamma ** (2))
-                else:
-                    lr = initial_lr
-            lr_each_step.append(lr)
-    return lr_each_step
+from src.lr_schedule import adjust_learning_rate
 
 def train(cfg):
 
@@ -107,10 +83,10 @@ def train(cfg):
                               warmup_epoch=cfg['warmup_epoch'])
 
     if cfg['optim'] == 'momentum':
-        opt = nn.Momentum(net.trainable_params(), lr, momentum)
+        opt = mindspore.nn.Momentum(net.trainable_params(), lr, momentum)
     elif cfg['optim'] == 'sgd':
-        opt = nn.SGD(params=net.trainable_params(), learning_rate=lr, momentum=momentum,
-                     weight_decay=weight_decay, loss_scale=1)
+        opt = mindspore.nn.SGD(params=net.trainable_params(), learning_rate=lr, momentum=momentum,
+                               weight_decay=weight_decay, loss_scale=1)
     else:
         raise ValueError('optim is not define.')
 
@@ -127,14 +103,13 @@ def train(cfg):
 
     print("============== Starting Training ==============")
     model.train(max_epoch, ds_train, callbacks=callback_list,
-                dataset_sink_mode=False)
-
+                dataset_sink_mode=True)
 
 
 if __name__ == '__main__':
 
-    setup_seed(1)
     config = cfg_res50
+    mindspore.common.seed.set_seed(config['seed'])
     print('train config:\n', config)
 
     train(cfg=config)