Cleanup & Evaluation now works

examples/run_squad.py

@@ -16,7 +16,7 @@
 """ Finetuning the library models for question-answering on SQuAD (DistilBERT, Bert, XLM, XLNet)."""
 
 from __future__ import absolute_import, division, print_function
-from transformers.data.processors.squad import SquadV1Processor
+from transformers.data.processors.squad import SquadV1Processor, SquadV2Processor
 
 import argparse
 import logging
@@ -45,9 +45,9 @@ from transformers import (WEIGHTS_NAME, BertConfig,
                                   XLNetTokenizer,
                                   DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer)
 
-from transformers import AdamW, get_linear_schedule_with_warmup, squad_convert_examples_to_features, read_squad_examples as sread_squad_examples
+from transformers import AdamW, get_linear_schedule_with_warmup, squad_convert_examples_to_features
 
-from utils_squad import (RawResult, write_predictions,
+from utils_squad import (convert_examples_to_features as old_convert, read_squad_examples as old_read, RawResult, write_predictions,
                          RawResultExtended, write_predictions_extended)
 
 # The follwing import is the official SQuAD evaluation script (2.0).
@@ -304,27 +304,19 @@ def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=Fal
         features = torch.load(cached_features_file)
     else:
         logger.info("Creating features from dataset file at %s", input_file)
-        examples = read_squad_examples(input_file=input_file,
-                                                is_training=not evaluate,
-                                                version_2_with_negative=args.version_2_with_negative)
-        keep_n_examples = 1000
-        processor = SquadV1Processor()
-        values = processor.get_dev_examples("examples/squad")
-        examples = values[:keep_n_examples]
-        features = squad_convert_examples_to_features(examples=exampless,
+
+        processor = SquadV2Processor()
+        examples = processor.get_dev_examples("examples/squad") if evaluate else processor.get_train_examples("examples/squad")
+        features = squad_convert_examples_to_features(
+            examples=examples,
             tokenizer=tokenizer,
             max_seq_length=args.max_seq_length,
             doc_stride=args.doc_stride,
             max_query_length=args.max_query_length,
             is_training=not evaluate,
-                                                cls_token_segment_id=2 if args.model_type in ['xlnet'] else 0,
-                                                pad_token_segment_id=3 if args.model_type in ['xlnet'] else 0,
-                                                cls_token_at_end=True if args.model_type in ['xlnet'] else False,
-                                                sequence_a_is_doc=True if args.model_type in ['xlnet'] else False)
-        print("DONE")
+            sequence_a_is_doc=True if args.model_type in ['xlnet'] else False
+        )
 
-        import sys
-        sys.exit()
 
         if args.local_rank in [-1, 0]:
             logger.info("Saving features into cached file %s", cached_features_file)
@@ -335,8 +327,8 @@ def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=Fal
 
     # Convert to Tensors and build dataset
     all_input_ids = torch.tensor([f.input_ids for f in features], dtype=torch.long)
-    all_input_mask = torch.tensor([f.input_mask for f in features], dtype=torch.long)
-    all_segment_ids = torch.tensor([f.segment_ids for f in features], dtype=torch.long)
+    all_input_mask = torch.tensor([f.attention_mask for f in features], dtype=torch.long)
+    all_segment_ids = torch.tensor([f.token_type_ids for f in features], dtype=torch.long)
     all_cls_index = torch.tensor([f.cls_index for f in features], dtype=torch.long)
     all_p_mask = torch.tensor([f.p_mask for f in features], dtype=torch.float)
     if evaluate:

transformers/data/processors/squad.py

@@ -74,26 +74,16 @@ def _is_whitespace(c):
 
 def squad_convert_examples_to_features(examples, tokenizer, max_seq_length,
                                        doc_stride, max_query_length, is_training,
-                                       cls_token_at_end=True,
-                                       cls_token='[CLS]', sep_token='[SEP]', pad_token=0,
-                                       sequence_a_segment_id=0, sequence_b_segment_id=1,
-                                       cls_token_segment_id=0, pad_token_segment_id=0,
-                                       mask_padding_with_zero=True,
                                        sequence_a_is_doc=False):
     """Loads a data file into a list of `InputBatch`s."""
 
-    cls_token = tokenizer.cls_token
-    sep_token = tokenizer.sep_token
-
     # Defining helper methods    
     unique_id = 1000000000
 
     features = []
-    new_features = []
     for (example_index, example) in enumerate(tqdm(examples)):
         if is_training and not example.is_impossible:
             # Get start and end position
-            answer_length = len(example.answer_text)
             start_position = example.start_position
             end_position = example.end_position
 
@@ -227,7 +217,7 @@ def squad_convert_examples_to_features(examples, tokenizer, max_seq_length,
                     end_position = tok_end_position - doc_start + doc_offset
 
 
-            new_features.append(NewSquadFeatures(
+            features.append(NewSquadFeatures(
                 span['input_ids'],
                 span['attention_mask'],
                 span['token_type_ids'],
@@ -247,7 +237,7 @@ def squad_convert_examples_to_features(examples, tokenizer, max_seq_length,
 
             unique_id += 1
 
-    return new_features
+    return features
 
 
 class SquadProcessor(DataProcessor):