!6503 [VM][Quant]Add discription of script parameters in model zoo of quantative network

Merge pull request !6503 from chenfei_mindspore/master

model_zoo/official/cv/mobilenetv2_quant/Readme.md

@@ -8,6 +8,7 @@
 - [Environment Requirements](#environment-requirements)
 - [Script Description](#script-description)
     - [Script and Sample Code](#script-and-sample-code)
+    - [Script Parameters](#script-parameters)
     - [Training Process](#training-process)
     - [Evaluation Process](#evaluation-process)
 - [Model Description](#model-description)
@@ -83,6 +84,31 @@ For FP16 operators, if the input data type is FP32, the backend of MindSpore wil
   ├── export.py     # export checkpoint files into air/onnx
 ```
 
+
+## [Script Parameters](#contents)
+
+Parameters for both training and evaluation can be set in config.py
+
+- config for MobileNetV2-quant, ImageNet2012 dataset
+
+  ```python
+  'class_num': 1000         # the number of classes in the dataset
+  'batch_size': 134         # training batch size
+  'epoch_size': 60         # training epochs of mobilenetv2-quant
+  'start epoch':200         # pretraining epochs of unquantative network
+  'warmup_epochs': 0        # number of warmup epochs
+  'lr': 0.3                 #learning rate
+  'momentum': 0.9           # momentum
+  'weight_decay': 4e-5      # weight decay value
+  'loss_scale': 1024        # the initial loss_scale value
+  'label_smooth': 0.1       #label smooth factor
+  'loss_scale': 1024        # the initial loss_scale value
+  'save_checkpoint':True    # whether save checkpoint file after training finish
+  'save_checkpoint_epochs': 1 # the step from which start to save checkpoint file.
+  'keep_checkpoint_max': 300  #  only keep the last keep_checkpoint_max checkpoint
+  'save_checkpoint_path': './checkpoint'  # the absolute full path to save the checkpoint file
+  ```
+
 ## [Training process](#contents)
 
 ### Usage

model_zoo/official/cv/mobilenetv2_quant/src/config.py

@@ -39,8 +39,6 @@ config_ascend_quant = ed({
 
 config_gpu_quant = ed({
     "num_classes": 1000,
-    "image_height": 224,
-    "image_width": 224,
     "batch_size": 134,
     "epoch_size": 60,
     "start_epoch": 200,
@@ -54,5 +52,4 @@ config_gpu_quant = ed({
     "save_checkpoint_epochs": 1,
     "keep_checkpoint_max": 300,
     "save_checkpoint_path": "./checkpoint",
-    "quantization_aware": True,
 })

model_zoo/official/cv/mobilenetv2_quant/src/dataset.py

@@ -64,7 +64,7 @@ def create_dataset(dataset_path, do_train, config, device_target, repeat_num=1,
     else:
         raise ValueError("Unsupported device_target.")
 
-    resize_height = config.image_height
+    resize_height = 224
 
     if do_train:
         buffer_size = 20480
@@ -128,7 +128,7 @@ def create_dataset_py(dataset_path, do_train, config, device_target, repeat_num=
     else:
         raise ValueError("Unsupported device target.")
 
-    resize_height = config.image_height
+    resize_height = 224
 
     if do_train:
         buffer_size = 20480

model_zoo/official/cv/resnet50_quant/Readme.md

@@ -8,6 +8,7 @@
 - [Environment Requirements](#environment-requirements)
 - [Script Description](#script-description)
     - [Script and Sample Code](#script-and-sample-code)
+    - [Script Parameters](#script-parameters)
     - [Training Process](#training-process)
     - [Evaluation Process](#evaluation-process)
 - [Model Description](#model-description)
@@ -83,6 +84,33 @@ For FP16 operators, if the input data type is FP32, the backend of MindSpore wil
 
 ```
 
+## [Script Parameters](#contents)
+
+Parameters for both training and evaluation can be set in config.py
+
+- config for Resnet50-quant, ImageNet2012 dataset
+
+  ```python
+  'class_num': 10           # the number of classes in the dataset
+  'batch_size': 32          # training batch size
+  'loss_scale': 1024        # the initial loss_scale value
+  'momentum': 0.9           # momentum
+  'weight_decay': 1e-4      # weight decay value
+  'epoch_size': 120         # total training epochs
+  'pretrained_epoch_size': 90   # pretraining epochs of resnet50, which is unquantative network of resnet50_quant
+  'data_load_mode': 'mindata' # the style of loading data into device
+  'save_checkpoint':True    # whether save checkpoint file after training finish
+  'save_checkpoint_epochs': 1 # the step from which start to save checkpoint file.
+  'keep_checkpoint_max': 50  #  only keep the last keep_checkpoint_max checkpoint
+  'save_checkpoint_path': './'  # the absolute full path to save the checkpoint file
+  "warmup_epochs": 0        # number of warmup epochs
+  'lr_decay_mode': "cosine" #learning rate decay mode, including steps, steps_decay, cosine or liner
+  'use_label_smooth': True  #whether use label smooth
+  'label_smooth_factor': 0.1 #label smooth factor
+  'lr_init': 0              # initial learning rate
+  'lr_max': 0.005           # the max learning rate
+  ```
+
 ## [Training process](#contents)
 
 ### Usage

model_zoo/official/cv/resnet50_quant/src/config.py

@@ -24,9 +24,6 @@ config_quant = ed({
     "weight_decay": 1e-4,
     "epoch_size": 120,
     "pretrained_epoch_size": 90,
-    "buffer_size": 1000,
-    "image_height": 224,
-    "image_width": 224,
     "data_load_mode": "mindata",
     "save_checkpoint": True,
     "save_checkpoint_epochs": 1,

model_zoo/official/cv/resnet50_quant/src/dataset.py

@@ -62,7 +62,7 @@ def create_dataset(dataset_path, do_train, repeat_num=1, batch_size=32, target="
         ds = load_func(num_parallel_workers=8, shuffle=True,
                        num_shards=device_num, shard_id=rank_id)
 
-    image_size = config.image_height
+    image_size = 224
     mean = [0.485 * 255, 0.456 * 255, 0.406 * 255]
     std = [0.229 * 255, 0.224 * 255, 0.225 * 255]
 

tests/st/quantization/lenet_quant/test_lenet_quant.py

@@ -47,7 +47,7 @@ def train_lenet():
     time_cb = TimeMonitor(data_size=ds_train.get_dataset_size())
     config_ck = CheckpointConfig(save_checkpoint_steps=cfg.save_checkpoint_steps,
                                  keep_checkpoint_max=cfg.keep_checkpoint_max)
-    ckpoint_cb = ModelCheckpoint(prefix="checkpoint_lenet", config=config_ck)
+    ckpoint_cb = ModelCheckpoint(prefix="ckpt_lenet_noquant", config=config_ck)
     model = Model(network, net_loss, net_opt, metrics={"Accuracy": Accuracy()})
 
     print("============== Starting Training Lenet==============")
@@ -58,7 +58,7 @@ def train_lenet():
 def train_lenet_quant():
     context.set_context(mode=context.GRAPH_MODE, device_target=device_target)
     cfg = quant_cfg
-    ckpt_path = './checkpoint_lenet-10_1875.ckpt'
+    ckpt_path = './ckpt_lenet_noquant-10_1875.ckpt'
     ds_train = create_dataset(os.path.join(data_path, "train"), cfg.batch_size, 1)
     step_size = ds_train.get_dataset_size()
 
@@ -81,7 +81,7 @@ def train_lenet_quant():
     # call back and monitor
     config_ckpt = CheckpointConfig(save_checkpoint_steps=cfg.epoch_size * step_size,
                                    keep_checkpoint_max=cfg.keep_checkpoint_max)
-    ckpt_callback = ModelCheckpoint(prefix="checkpoint_lenet", config=config_ckpt)
+    ckpt_callback = ModelCheckpoint(prefix="ckpt_lenet_quant", config=config_ckpt)
 
     # define model
     model = Model(network, net_loss, net_opt, metrics={"Accuracy": Accuracy()})
@@ -96,7 +96,7 @@ def eval_quant():
     context.set_context(mode=context.GRAPH_MODE, device_target=device_target)
     cfg = quant_cfg
     ds_eval = create_dataset(os.path.join(data_path, "test"), cfg.batch_size, 1)
-    ckpt_path = './checkpoint_lenet_1-10_937.ckpt'
+    ckpt_path = './ckpt_lenet_quant-10_937.ckpt'
     # define fusion network
     network = LeNet5Fusion(cfg.num_classes)
     # convert fusion network to quantization aware network