modify config param

example/resnet101_imagenet2012/README.md

@@ -51,8 +51,8 @@ Parameters for both training and evaluating can be set in config.py.
 "image_height": 224,              # image height
 "image_width": 224,               # image width
 "save_checkpoint": True,          # whether save checkpoint or not
-"save_checkpoint_steps": 500,     # the step interval between two checkpoints. By default, the last checkpoint will be saved after the last step
-"keep_checkpoint_max": 40,        # only keep the last keep_checkpoint_max checkpoint
+"save_checkpoint_epochs": 1,      # the epoch interval between two checkpoints. By default, the last checkpoint will be saved after the last epoch
+"keep_checkpoint_max": 10,        # only keep the last keep_checkpoint_max checkpoint
 "save_checkpoint_path": "./",     # path to save checkpoint relative to the executed path
 "warmup_epochs": 0,               # number of warmup epoch
 "lr_decay_mode": "cosine"         # decay mode for generating learning rate

example/resnet101_imagenet2012/config.py

@@ -28,8 +28,8 @@ config = ed({
     "image_height": 224,
     "image_width": 224,
     "save_checkpoint": True,
-    "save_checkpoint_steps": 500,
-    "keep_checkpoint_max": 40,
+    "save_checkpoint_epochs": 1,
+    "keep_checkpoint_max": 10,
     "save_checkpoint_path": "./",
     "warmup_epochs": 0,
     "lr_decay_mode": "cosine",

example/resnet101_imagenet2012/eval.py

@@ -54,7 +54,7 @@ if __name__ == '__main__':
     if not args_opt.do_eval and args_opt.run_distribute:
         context.set_auto_parallel_context(device_num=args_opt.device_num, parallel_mode=ParallelMode.DATA_PARALLEL,
                                           mirror_mean=True, parameter_broadcast=True)
-        auto_parallel_context().set_all_reduce_fusion_split_indices([140])
+        auto_parallel_context().set_all_reduce_fusion_split_indices([180, 313])
         init()
 
     epoch_size = config.epoch_size

example/resnet101_imagenet2012/train.py

@@ -59,7 +59,7 @@ if __name__ == '__main__':
     if not args_opt.do_eval and args_opt.run_distribute:
         context.set_auto_parallel_context(device_num=args_opt.device_num, parallel_mode=ParallelMode.DATA_PARALLEL,
                                           mirror_mean=True, parameter_broadcast=True)
-        auto_parallel_context().set_all_reduce_fusion_split_indices([140])
+        auto_parallel_context().set_all_reduce_fusion_split_indices([180, 313])
         init()
 
     epoch_size = config.epoch_size
@@ -91,7 +91,7 @@ if __name__ == '__main__':
         loss_cb = LossMonitor()
         cb = [time_cb, loss_cb]
         if config.save_checkpoint:
-            config_ck = CheckpointConfig(save_checkpoint_steps=config.save_checkpoint_steps,
+            config_ck = CheckpointConfig(save_checkpoint_steps=config.save_checkpoint_epochs*step_size,
                                          keep_checkpoint_max=config.keep_checkpoint_max)
             ckpt_cb = ModelCheckpoint(prefix="resnet", directory=config.save_checkpoint_path, config=config_ck)
             cb += [ckpt_cb]