Merge pull request #668 from jeonsworld/patch-2

apply Whole Word Masking technique

examples/lm_finetuning/pregenerate_training_data.py

@@ -4,11 +4,11 @@ from tqdm import tqdm, trange
 from tempfile import TemporaryDirectory
 import shelve
 
-from random import random, randrange, randint, shuffle, choice, sample
+from random import random, randrange, randint, shuffle, choice
 from pytorch_pretrained_bert.tokenization import BertTokenizer
 import numpy as np
 import json
-
+import collections
 
 class DocumentDatabase:
     def __init__(self, reduce_memory=False):
@@ -98,42 +98,77 @@ def truncate_seq_pair(tokens_a, tokens_b, max_num_tokens):
         else:
             trunc_tokens.pop()
 
+MaskedLmInstance = collections.namedtuple("MaskedLmInstance",
+                                          ["index", "label"])
 
-def create_masked_lm_predictions(tokens, masked_lm_prob, max_predictions_per_seq, vocab_list):
+def create_masked_lm_predictions(tokens, masked_lm_prob, max_predictions_per_seq, whole_word_mask, vocab_list):
     """Creates the predictions for the masked LM objective. This is mostly copied from the Google BERT repo, but
     with several refactors to clean it up and remove a lot of unnecessary variables."""
     cand_indices = []
     for (i, token) in enumerate(tokens):
         if token == "[CLS]" or token == "[SEP]":
             continue
-        cand_indices.append(i)
+        # Whole Word Masking means that if we mask all of the wordpieces
+        # corresponding to an original word. When a word has been split into
+        # WordPieces, the first token does not have any marker and any subsequence
+        # tokens are prefixed with ##. So whenever we see the ## token, we
+        # append it to the previous set of word indexes.
+        #
+        # Note that Whole Word Masking does *not* change the training code
+        # at all -- we still predict each WordPiece independently, softmaxed
+        # over the entire vocabulary.
+        if (whole_word_mask and len(cand_indices) >= 1 and token.startswith("##")):
+            cand_indices[-1].append(i)
+        else:
+            cand_indices.append([i])
 
     num_to_mask = min(max_predictions_per_seq,
                       max(1, int(round(len(tokens) * masked_lm_prob))))
     shuffle(cand_indices)
-    mask_indices = sorted(sample(cand_indices, num_to_mask))
-    masked_token_labels = []
-    for index in mask_indices:
-        # 80% of the time, replace with [MASK]
-        if random() < 0.8:
-            masked_token = "[MASK]"
-        else:
-            # 10% of the time, keep original
-            if random() < 0.5:
-                masked_token = tokens[index]
-            # 10% of the time, replace with random word
+    masked_lms = []
+    covered_indexes = set()
+    for index_set in cand_indices:
+        if len(masked_lms) >= num_to_mask:
+            break
+        # If adding a whole-word mask would exceed the maximum number of
+        # predictions, then just skip this candidate.
+        if len(masked_lms) + len(index_set) > num_to_mask:
+            continue
+        is_any_index_covered = False
+        for index in index_set:
+            if index in covered_indexes:
+                is_any_index_covered = True
+                break
+        if is_any_index_covered:
+            continue
+        for index in index_set:
+            covered_indexes.add(index)
+
+            masked_token = None
+            # 80% of the time, replace with [MASK]
+            if random() < 0.8:
+                masked_token = "[MASK]"
             else:
-                masked_token = choice(vocab_list)
-        masked_token_labels.append(tokens[index])
-        # Once we've saved the true label for that token, we can overwrite it with the masked version
-        tokens[index] = masked_token
+                # 10% of the time, keep original
+                if random() < 0.5:
+                    masked_token = tokens[index]
+                # 10% of the time, replace with random word
+                else:
+                    masked_token = choice(vocab_list)
+            masked_lms.append(MaskedLmInstance(index=index, label=tokens[index]))
+            tokens[index] = masked_token
+
+    assert len(masked_lms) <= num_to_mask
+    masked_lms = sorted(masked_lms, key=lambda x: x.index)
+    mask_indices = [p.index for p in masked_lms]
+    masked_token_labels = [p.label for p in masked_lms]
 
     return tokens, mask_indices, masked_token_labels
 
 
 def create_instances_from_document(
         doc_database, doc_idx, max_seq_length, short_seq_prob,
-        masked_lm_prob, max_predictions_per_seq, vocab_list):
+        masked_lm_prob, max_predictions_per_seq, whole_word_mask, vocab_list):
     """This code is mostly a duplicate of the equivalent function from Google BERT's repo.
     However, we make some changes and improvements. Sampling is improved and no longer requires a loop in this function.
     Also, documents are sampled proportionally to the number of sentences they contain, which means each sentence
@@ -213,7 +248,7 @@ def create_instances_from_document(
                 segment_ids = [0 for _ in range(len(tokens_a) + 2)] + [1 for _ in range(len(tokens_b) + 1)]
 
                 tokens, masked_lm_positions, masked_lm_labels = create_masked_lm_predictions(
-                    tokens, masked_lm_prob, max_predictions_per_seq, vocab_list)
+                    tokens, masked_lm_prob, max_predictions_per_seq, whole_word_mask, vocab_list)
 
                 instance = {
                     "tokens": tokens,
@@ -237,7 +272,8 @@ def main():
                         choices=["bert-base-uncased", "bert-large-uncased", "bert-base-cased",
                                  "bert-base-multilingual", "bert-base-chinese"])
     parser.add_argument("--do_lower_case", action="store_true")
-
+    parser.add_argument("--do_whole_word_mask", action="store_true",
+                        help="Whether to use whole word masking rather than per-WordPiece masking.")
     parser.add_argument("--reduce_memory", action="store_true",
                         help="Reduce memory usage for large datasets by keeping data on disc rather than in memory")
 
@@ -284,7 +320,7 @@ def main():
                     doc_instances = create_instances_from_document(
                         docs, doc_idx, max_seq_length=args.max_seq_len, short_seq_prob=args.short_seq_prob,
                         masked_lm_prob=args.masked_lm_prob, max_predictions_per_seq=args.max_predictions_per_seq,
-                        vocab_list=vocab_list)
+                        whole_word_mask=args.do_whole_word_mask, vocab_list=vocab_list)
                     doc_instances = [json.dumps(instance) for instance in doc_instances]
                     for instance in doc_instances:
                         epoch_file.write(instance + '\n')