refactor: adapt to static graph in deprecating MixCELoss

ppcls/engine/engine.py

@@ -112,6 +112,7 @@ class Engine(object):
             }
             paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING)
 
+        #TODO(gaotingquan): support rec
         class_num = config["Arch"].get("class_num", None)
         self.config["DataLoader"].update({"class_num": class_num})
         # build dataloader

ppcls/static/program.py

@@ -41,13 +41,14 @@ from ppcls.utils.misc import AverageMeter
 from ppcls.utils import logger, profiler
 
 
-def create_feeds(image_shape, use_mix=None, dtype="float32"):
+def create_feeds(image_shape, use_mix=False, class_num=None, dtype="float32"):
     """
     Create feeds as model input
 
     Args:
         image_shape(list[int]): model input shape, such as [3, 224, 224]
         use_mix(bool): whether to use mix(include mixup, cutmix, fmix)
+        class_num(int): the class number of network, required if use_mix
 
     Returns:
         feeds(dict): dict of model input variables
@@ -55,13 +56,14 @@ def create_feeds(image_shape, use_mix=None, dtype="float32"):
     feeds = OrderedDict()
     feeds['data'] = paddle.static.data(
         name="data", shape=[None] + image_shape, dtype=dtype)
+
     if use_mix:
-        feeds['y_a'] = paddle.static.data(
-            name="y_a", shape=[None, 1], dtype="int64")
-        feeds['y_b'] = paddle.static.data(
-            name="y_b", shape=[None, 1], dtype="int64")
-        feeds['lam'] = paddle.static.data(
-            name="lam", shape=[None, 1], dtype=dtype)
+        if class_num is None:
+            msg = "When use MixUp, CutMix and so on, you must set class_num."
+            logger.error(msg)
+            raise Exception(msg)
+        feeds['target'] = paddle.static.data(
+            name="target", shape=[None, class_num], dtype="float32")
     else:
         feeds['label'] = paddle.static.data(
             name="label", shape=[None, 1], dtype="int64")
@@ -74,6 +76,7 @@ def create_fetchs(out,
                   architecture,
                   topk=5,
                   epsilon=None,
+                  class_num=None,
                   use_mix=False,
                   config=None,
                   mode="Train"):
@@ -88,6 +91,7 @@ def create_fetchs(out,
             name(such as ResNet50) is needed
         topk(int): usually top5
         epsilon(float): parameter for label smoothing, 0.0 <= epsilon <= 1.0
+        class_num(int): the class number of network, required if use_mix
         use_mix(bool): whether to use mix(include mixup, cutmix, fmix)
         config(dict): model config
 
@@ -97,18 +101,16 @@ def create_fetchs(out,
     fetchs = OrderedDict()
     # build loss
     if use_mix:
-        y_a = paddle.reshape(feeds['y_a'], [-1, 1])
-        y_b = paddle.reshape(feeds['y_b'], [-1, 1])
-        lam = paddle.reshape(feeds['lam'], [-1, 1])
+        if class_num is None:
+            msg = "When use MixUp, CutMix and so on, you must set class_num."
+            logger.error(msg)
+            raise Exception(msg)
+        target = paddle.reshape(feeds['target'], [-1, class_num])
     else:
         target = paddle.reshape(feeds['label'], [-1, 1])
 
     loss_func = build_loss(config["Loss"][mode])
-
-    if use_mix:
-        loss_dict = loss_func(out, [y_a, y_b, lam])
-    else:
-        loss_dict = loss_func(out, target)
+    loss_dict = loss_func(out, target)
 
     loss_out = loss_dict["loss"]
     fetchs['loss'] = (loss_out, AverageMeter('loss', '7.4f', need_avg=True))
@@ -218,6 +220,7 @@ def mixed_precision_optimizer(config, optimizer):
 def build(config,
           main_prog,
           startup_prog,
+          class_num=None,
           step_each_epoch=100,
           is_train=True,
           is_distributed=True):
@@ -233,6 +236,7 @@ def build(config,
         config(dict): config
         main_prog(): main program
         startup_prog(): startup program
+        class_num(int): the class number of network, required if use_mix
         is_train(bool): train or eval
         is_distributed(bool): whether to use distributed training method
 
@@ -245,10 +249,10 @@ def build(config,
             mode = "Train" if is_train else "Eval"
             use_mix = "batch_transform_ops" in config["DataLoader"][mode][
                 "dataset"]
-            use_dali = config["Global"].get('use_dali', False)
             feeds = create_feeds(
                 config["Global"]["image_shape"],
-                use_mix=use_mix,
+                use_mix,
+                class_num=class_num,
                 dtype="float32")
 
             # build model
@@ -264,6 +268,7 @@ def build(config,
                 feeds,
                 config["Arch"],
                 epsilon=config.get('ls_epsilon'),
+                class_num=class_num,
                 use_mix=use_mix,
                 config=config,
                 mode=mode)

ppcls/static/train.py

@@ -112,6 +112,8 @@ def main(args):
     eval_dataloader = None
     use_dali = global_config.get('use_dali', False)
 
+    class_num = config["Arch"].get("class_num", None)
+    config["DataLoader"].update({"class_num": class_num})
     train_dataloader = build_dataloader(
         config["DataLoader"], "Train", device=device, use_dali=use_dali)
     if global_config["eval_during_train"]:
@@ -131,6 +133,7 @@ def main(args):
         config,
         train_prog,
         startup_prog,
+        class_num,
         step_each_epoch=step_each_epoch,
         is_train=True,
         is_distributed=global_config.get("is_distributed", True))