Update chnsenticorp examples for qianyan dataset modification(#485)

Author: Steffy-zxf

examples/sentiment_analysis/skep/train_sentence.py

@@ -172,13 +172,12 @@ def create_dataloader(dataset,
     set_seed(args.seed)
     if args.model_name == "skep_ernie_1.0_large_ch":
         dataset_name = "chnsenticorp"
-        train_ds, dev_ds, test_ds = load_dataset(
-            dataset_name, splits=["train", "dev", "test"])
+        train_ds, dev_ds = load_dataset(dataset_name, splits=["train", "dev"])
 
     else:
         dataset_name = "sst-2"
-        train_ds, dev_ds, test_ds = load_dataset(
-            "glue", dataset_name, splits=["train", "dev", "test"])
+        train_ds, dev_ds = load_dataset(
+            "glue", dataset_name, splits=["train", "dev"])
     label_map = {0: 'negative', 1: 'positive'}
 
     model = SkepForSequenceClassification.from_pretrained(

examples/text_classification/pretrained_models/train.py

@@ -152,8 +152,7 @@ def do_train():
 
     set_seed(args.seed)
 
-    train_ds, dev_ds, test_ds = load_dataset(
-        "chnsenticorp", splits=["train", "dev", "test"])
+    train_ds, dev_ds = load_dataset("chnsenticorp", splits=["train", "dev"])
 
     # If you wanna use bert/roberta/electra pretrained model,
     # model = ppnlp.transformers.BertForSequenceClassification.from_pretrained('bert-base-chinese', num_class=2)
@@ -191,12 +190,6 @@ def do_train():
         batch_size=args.batch_size,
         batchify_fn=batchify_fn,
         trans_fn=trans_func)
-    test_data_loader = create_dataloader(
-        test_ds,
-        mode='test',
-        batch_size=args.batch_size,
-        batchify_fn=batchify_fn,
-        trans_fn=trans_func)
 
     if args.init_from_ckpt and os.path.isfile(args.init_from_ckpt):
         state_dict = paddle.load(args.init_from_ckpt)
@@ -254,10 +247,6 @@ def do_train():
                 model._layers.save_pretrained(save_dir)
                 tokenizer.save_pretrained(save_dir)
 
-    if rank == 0:
-        print('Evaluating on test data.')
-        evaluate(model, criterion, metric, test_data_loader)
-
 
 if __name__ == "__main__":
     do_train()

examples/text_classification/rnn/train.py

@@ -94,8 +94,7 @@ def create_dataloader(dataset,
     vocab = Vocab.load_vocabulary(
         args.vocab_path, unk_token='[UNK]', pad_token='[PAD]')
     # Loads dataset.
-    train_ds, dev_ds, test_ds = load_dataset(
-        "chnsenticorp", splits=["train", "dev", "test"])
+    train_ds, dev_ds = load_dataset("chnsenticorp", splits=["train", "dev"])
 
     # Constructs the newtork.
     network = args.network.lower()
@@ -180,12 +179,6 @@ def create_dataloader(dataset,
         batch_size=args.batch_size,
         mode='validation',
         batchify_fn=batchify_fn)
-    test_loader = create_dataloader(
-        test_ds,
-        trans_fn=trans_fn,
-        batch_size=args.batch_size,
-        mode='test',
-        batchify_fn=batchify_fn)
 
     optimizer = paddle.optimizer.Adam(
         parameters=model.parameters(), learning_rate=args.lr)
@@ -208,7 +201,3 @@ def create_dataloader(dataset,
               epochs=args.epochs,
               save_dir=args.save_dir,
               callbacks=callback)
-
-    # Finally tests model.
-    results = model.evaluate(test_loader)
-    print("Finally test acc: %.5f" % results['acc'])

examples/word_embedding/train.py

@@ -151,8 +151,7 @@ def forward(self, text, seq_len=None):
     if '[PAD]' not in vocab:
         vocab['[PAD]'] = len(vocab)
     # Loads dataset.
-    train_ds, dev_ds, test_ds = load_dataset(
-        "chnsenticorp", splits=["train", "dev", "test"])
+    train_ds, dev_ds = load_dataset("chnsenticorp", splits=["train", "dev"])
 
     # Constructs the newtork.
     model = BoWModel(
@@ -187,12 +186,6 @@ def forward(self, text, seq_len=None):
         batch_size=args.batch_size,
         mode='validation',
         pad_token_id=vocab['[PAD]'])
-    test_loader = create_dataloader(
-        test_ds,
-        trans_fn=trans_fn,
-        batch_size=args.batch_size,
-        mode='test',
-        pad_token_id=vocab['[PAD]'])
 
     optimizer = paddle.optimizer.Adam(
         parameters=model.parameters(), learning_rate=args.lr)
@@ -219,7 +212,3 @@ def forward(self, text, seq_len=None):
               epochs=args.epochs,
               save_dir=args.save_dir,
               callbacks=callback)
-
-    # Finally tests model.
-    results = model.evaluate(test_loader, callbacks=callback)
-    print("Finally test acc: %.5f" % results['acc'])