update finetune task api name for simplication

demo/ernie-classification/ernie_tiny_demo.py

@@ -20,11 +20,7 @@ with fluid.program_guard(program):
         feature=pooled_output, label=label, num_classes=dataset.num_labels)
 
 # Step4
-strategy = hub.AdamWeightDecayStrategy(
-    learning_rate=5e-5,
-    warmup_proportion=0.1,
-    warmup_strategy="linear_warmup_decay",
-    weight_decay=0.01)
+strategy = hub.AdamWeightDecayStrategy(learning_rate=5e-5, weight_decay=0.01)
 
 config = hub.RunConfig(
     use_cuda=True, num_epoch=3, batch_size=32, strategy=strategy)

demo/ernie-classification/run_sentiment_cls.sh

@@ -3,6 +3,7 @@ export CUDA_VISIBLE_DEVICES=5
 CKPT_DIR="./ckpt_sentiment_cls"
 python -u sentiment_cls.py \
                    --batch_size 32 \
+                   --use_gpu=False \
                    --weight_decay  0.01 \
                    --checkpoint_dir $CKPT_DIR \
                    --num_epoch 3 \

demo/ernie-classification/sentiment_cls.py

@@ -14,6 +14,7 @@
 """Finetuning on classification task """
 
 import argparse
+import ast
 
 import paddle.fluid as fluid
 import paddlehub as hub
@@ -21,6 +22,7 @@ import paddlehub as hub
 # yapf: disable
 parser = argparse.ArgumentParser(__doc__)
 parser.add_argument("--num_epoch", type=int, default=3, help="Number of epoches for fine-tuning.")
+parser.add_argument("--use_gpu", type=ast.literal_eval, default=False, help="Whether use GPU for finetuning, input should be True or False")
 parser.add_argument("--learning_rate", type=float, default=5e-5, help="Learning rate used to train with warmup.")
 parser.add_argument("--weight_decay", type=float, default=0.01, help="Weight decay rate for L2 regularizer.")
 parser.add_argument("--data_dir", type=str, default=None, help="Path to training data.")
@@ -70,8 +72,9 @@ if __name__ == '__main__':
         )
 
         # Setup runing config for PaddleHub Finetune API
+        print(args.use_gpu)
         config = hub.RunConfig(
-            use_cuda=True,
+            use_cuda=args.use_gpu,
             num_epoch=args.num_epoch,
             batch_size=args.batch_size,
             checkpoint_dir=args.checkpoint_dir,

demo/image-classification/retrain.py

-import paddle
 import paddle.fluid as fluid
-
 import paddlehub as hub
 
-
-def train():
+if __name__ == "__main__":
     resnet_module = hub.Module(module_dir="ResNet50.hub_module")
-    input_dict, output_dict, program = resnet_module.context(
-        sign_name="feature_map", trainable=True)
+    input_dict, output_dict, program = resnet_module.context(trainable=True)
     dataset = hub.dataset.Flowers()
     data_reader = hub.reader.ImageClassificationReader(
         image_width=resnet_module.get_excepted_image_width(),
@@ -29,11 +25,7 @@ def train():
 
         feed_list = [img.name, label.name]
 
-        task = hub.create_img_classification_task(
+        task = hub.create_img_cls_task(
             feature=feature_map, label=label, num_classes=dataset.num_labels)
         hub.finetune_and_eval(
             task, feed_list=feed_list, data_reader=data_reader, config=config)
-
-
-if __name__ == "__main__":
-    train()

paddlehub/__init__.py

@@ -34,9 +34,9 @@ from .module.manager import default_module_manager
 from .io.type import DataType
 
 from .finetune.task import Task
-from .finetune.task import create_seq_labeling_task
-from .finetune.task import create_text_classification_task
-from .finetune.task import create_img_classification_task
+from .finetune.task import create_seq_label_task
+from .finetune.task import create_text_cls_task
+from .finetune.task import create_img_cls_task
 from .finetune.finetune import finetune_and_eval
 from .finetune.config import RunConfig
 from .finetune.strategy import AdamWeightDecayStrategy

paddlehub/finetune/finetune.py

@@ -45,10 +45,10 @@ def _do_memory_optimization(task, config):
         time_used = time.time() - optimize_time_begin
         logger.info("Memory optimization done! Time elapsed %f sec" % time_used)
 
-    lower_mem, upper_mem, unit = fluid.contrib.memory_usage(
-        program=fluid.default_main_program(), batch_size=config.batch_size)
-    logger.info("Theoretical memory usage in training: %.2f - %.2f %s" %
-                (lower_mem, upper_mem, unit)),
+    # lower_mem, upper_mem, unit = fluid.contrib.memory_usage(
+    #     program=task.main_program(), batch_size=config.batch_size)
+    # logger.info("Theoretical memory usage in training: %.2f - %.2f %s" %
+    #             (lower_mem, upper_mem, unit)),
 
 
 def _finetune_seq_label_task(task,

paddlehub/finetune/task.py

@@ -62,10 +62,7 @@ class Task(object):
         return metric_variable_names
 
 
-def create_text_classification_task(feature,
-                                    label,
-                                    num_classes,
-                                    hidden_units=None):
+def create_text_cls_task(feature, label, num_classes, hidden_units=None):
     """
     Append a multi-layer perceptron classifier for binary classification base
     on input feature
@@ -108,10 +105,7 @@ def create_text_classification_task(feature,
     return task
 
 
-def create_img_classification_task(feature,
-                                   label,
-                                   num_classes,
-                                   hidden_units=None):
+def create_img_cls_task(feature, label, num_classes, hidden_units=None):
     """
     Create the transfer learning task for image classification.
     Args:
@@ -153,13 +147,13 @@ def create_img_classification_task(feature,
         "num_example": num_example
     }
 
-    task = Task("text_classification", graph_var_dict,
+    task = Task("image_classification", graph_var_dict,
                 fluid.default_main_program(), fluid.default_startup_program())
 
     return task
 
 
-def create_seq_labeling_task(feature, labels, seq_len, num_classes):
+def create_seq_label_task(feature, labels, seq_len, num_classes):
     logits = fluid.layers.fc(
         input=feature,
         size=num_classes,