remove padding warning (#916)

Author: eric-haibin-lin

scripts/bert/finetune_classifier.py

@@ -295,9 +295,12 @@ def preprocess_data(tokenizer, task, batch_size, dev_batch_size, max_len, pad=Fa
     data_train_len = data_train.transform(
         lambda input_id, length, segment_id, label_id: length, lazy=False)
     # bucket sampler for training
+    pad_val = vocabulary[vocabulary.padding_token]
     batchify_fn = nlp.data.batchify.Tuple(
-        nlp.data.batchify.Pad(axis=0), nlp.data.batchify.Stack(),
-        nlp.data.batchify.Pad(axis=0), nlp.data.batchify.Stack(label_dtype))
+        nlp.data.batchify.Pad(axis=0, pad_val=pad_val), # input
+        nlp.data.batchify.Stack(),                      # length
+        nlp.data.batchify.Pad(axis=0, pad_val=0),       # segment
+        nlp.data.batchify.Stack(label_dtype))           # label
     batch_sampler = nlp.data.sampler.FixedBucketSampler(
         data_train_len,
         batch_size=batch_size,
@@ -327,8 +330,8 @@ def preprocess_data(tokenizer, task, batch_size, dev_batch_size, max_len, pad=Fa
 
     # batchify for data test
     test_batchify_fn = nlp.data.batchify.Tuple(
-        nlp.data.batchify.Pad(axis=0), nlp.data.batchify.Stack(),
-        nlp.data.batchify.Pad(axis=0))
+        nlp.data.batchify.Pad(axis=0, pad_val=pad_val), nlp.data.batchify.Stack(),
+        nlp.data.batchify.Pad(axis=0, pad_val=0))
     # transform for data test
     test_trans = BERTDatasetTransform(tokenizer, max_len,
                                       class_labels=None,