def __repr__(self): s = "" s += "qas_id: %s" % (tokenization.printable_text(self.qas_id)) s += ", question_text: %s" % (tokenization.printable_text( self.question_text)) s += ", doc_tokens: [%s]" % (" ".join(self.doc_tokens)) if self.start_position: s += ", start_position: %d" % (self.start_position) if self.start_position: s += ", end_position: %d" % (self.end_position) if self.start_position: s += ", is_impossible: %r" % (self.is_impossible) return s
def convert_single_example(ex_index, example, label_list, max_seq_length, tokenizer): """Converts a single `InputExample` into a single `InputFeatures`.""" label_map = {} for (i, label) in enumerate(label_list): label_map[label] = i tokens_a = tokenizer.tokenize(example.text_a) tokens_b = None if example.text_b: tokens_b = tokenizer.tokenize(example.text_b) if tokens_b: # Modifies `tokens_a` and `tokens_b` in place so that the total # length is less than the specified length. # Account for [CLS], [SEP], [SEP] with "- 3" _truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3) else: # Account for [CLS] and [SEP] with "- 2" if len(tokens_a) > max_seq_length - 2: tokens_a = tokens_a[0:(max_seq_length - 2)] # The convention in BERT is: # (a) For sequence pairs: # tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP] # type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1 # (b) For single sequences: # tokens: [CLS] the dog is hairy . [SEP] # type_ids: 0 0 0 0 0 0 0 # # Where "type_ids" are used to indicate whether this is the first # sequence or the second sequence. The embedding vectors for `type=0` and # `type=1` were learned during pre-training and are added to the wordpiece # embedding vector (and position vector). This is not *strictly* necessary # since the [SEP] token unambiguously separates the sequences, but it makes # it easier for the model to learn the concept of sequences. # # For classification tasks, the first vector (corresponding to [CLS]) is # used as the "sentence vector". Note that this only makes sense because # the entire model is fine-tuned. tokens = [] segment_ids = [] tokens.append("[CLS]") segment_ids.append(0) for token in tokens_a: tokens.append(token) segment_ids.append(0) tokens.append("[SEP]") segment_ids.append(0) if tokens_b: for token in tokens_b: tokens.append(token) segment_ids.append(1) tokens.append("[SEP]") segment_ids.append(1) input_ids = tokenizer.convert_tokens_to_ids(tokens) # The mask has 1 for real tokens and 0 for padding tokens. Only real # tokens are attended to. input_mask = [1] * len(input_ids) # Zero-pad up to the sequence length. while len(input_ids) < max_seq_length: input_ids.append(0) input_mask.append(0) segment_ids.append(0) assert len(input_ids) == max_seq_length assert len(input_mask) == max_seq_length assert len(segment_ids) == max_seq_length label_id = label_map[example.label] if ex_index < 5: logging.info("*** Example ***") logging.info("guid: %s", (example.guid)) logging.info( "tokens: %s", " ".join([tokenization.printable_text(x) for x in tokens])) logging.info("input_ids: %s", " ".join([str(x) for x in input_ids])) logging.info("input_mask: %s", " ".join([str(x) for x in input_mask])) logging.info("segment_ids: %s", " ".join([str(x) for x in segment_ids])) logging.info("label: %s (id = %d)", example.label, label_id) feature = InputFeatures(input_ids=input_ids, input_mask=input_mask, segment_ids=segment_ids, label_id=label_id, is_real_example=True) return feature
def convert_examples_to_features(examples, tokenizer, max_seq_length, doc_stride, max_query_length, is_training, output_fn, batch_size=None): """Loads a data file into a list of `InputBatch`s.""" base_id = 1000000000 unique_id = base_id feature = None for (example_index, example) in enumerate(examples): query_tokens = tokenizer.tokenize(example.question_text) if len(query_tokens) > max_query_length: query_tokens = query_tokens[0:max_query_length] tok_to_orig_index = [] orig_to_tok_index = [] all_doc_tokens = [] for (i, token) in enumerate(example.doc_tokens): orig_to_tok_index.append(len(all_doc_tokens)) sub_tokens = tokenizer.tokenize(token) for sub_token in sub_tokens: tok_to_orig_index.append(i) all_doc_tokens.append(sub_token) tok_start_position = None tok_end_position = None if is_training and example.is_impossible: tok_start_position = -1 tok_end_position = -1 if is_training and not example.is_impossible: tok_start_position = orig_to_tok_index[example.start_position] if example.end_position < len(example.doc_tokens) - 1: tok_end_position = orig_to_tok_index[example.end_position + 1] - 1 else: tok_end_position = len(all_doc_tokens) - 1 (tok_start_position, tok_end_position) = _improve_answer_span( all_doc_tokens, tok_start_position, tok_end_position, tokenizer, example.orig_answer_text) # The -3 accounts for [CLS], [SEP] and [SEP] max_tokens_for_doc = max_seq_length - len(query_tokens) - 3 # We can have documents that are longer than the maximum sequence length. # To deal with this we do a sliding window approach, where we take chunks # of the up to our max length with a stride of `doc_stride`. _DocSpan = collections.namedtuple( # pylint: disable=invalid-name "DocSpan", ["start", "length"]) doc_spans = [] start_offset = 0 while start_offset < len(all_doc_tokens): length = len(all_doc_tokens) - start_offset if length > max_tokens_for_doc: length = max_tokens_for_doc doc_spans.append(_DocSpan(start=start_offset, length=length)) if start_offset + length == len(all_doc_tokens): break start_offset += min(length, doc_stride) for (doc_span_index, doc_span) in enumerate(doc_spans): tokens = [] token_to_orig_map = {} token_is_max_context = {} segment_ids = [] tokens.append("[CLS]") segment_ids.append(0) for token in query_tokens: tokens.append(token) segment_ids.append(0) tokens.append("[SEP]") segment_ids.append(0) for i in range(doc_span.length): split_token_index = doc_span.start + i token_to_orig_map[len( tokens)] = tok_to_orig_index[split_token_index] is_max_context = _check_is_max_context(doc_spans, doc_span_index, split_token_index) token_is_max_context[len(tokens)] = is_max_context tokens.append(all_doc_tokens[split_token_index]) segment_ids.append(1) tokens.append("[SEP]") segment_ids.append(1) input_ids = tokenizer.convert_tokens_to_ids(tokens) # The mask has 1 for real tokens and 0 for padding tokens. Only real # tokens are attended to. input_mask = [1] * len(input_ids) # Zero-pad up to the sequence length. while len(input_ids) < max_seq_length: input_ids.append(0) input_mask.append(0) segment_ids.append(0) assert len(input_ids) == max_seq_length assert len(input_mask) == max_seq_length assert len(segment_ids) == max_seq_length start_position = None end_position = None if is_training and not example.is_impossible: # For training, if our document chunk does not contain an annotation # we throw it out, since there is nothing to predict. doc_start = doc_span.start doc_end = doc_span.start + doc_span.length - 1 out_of_span = False if not (tok_start_position >= doc_start and tok_end_position <= doc_end): out_of_span = True if out_of_span: start_position = 0 end_position = 0 else: doc_offset = len(query_tokens) + 2 start_position = tok_start_position - doc_start + doc_offset end_position = tok_end_position - doc_start + doc_offset if is_training and example.is_impossible: start_position = 0 end_position = 0 if example_index < 20: logging.info("*** Example ***") logging.info("unique_id: %s", (unique_id)) logging.info("example_index: %s", (example_index)) logging.info("doc_span_index: %s", (doc_span_index)) logging.info( "tokens: %s", " ".join([tokenization.printable_text(x) for x in tokens])) logging.info( "token_to_orig_map: %s", " ".join([ "%d:%d" % (x, y) for (x, y) in six.iteritems(token_to_orig_map) ])) logging.info( "token_is_max_context: %s", " ".join([ "%d:%s" % (x, y) for (x, y) in six.iteritems(token_is_max_context) ])) logging.info("input_ids: %s", " ".join([str(x) for x in input_ids])) logging.info("input_mask: %s", " ".join([str(x) for x in input_mask])) logging.info("segment_ids: %s", " ".join([str(x) for x in segment_ids])) if is_training and example.is_impossible: logging.info("impossible example") if is_training and not example.is_impossible: answer_text = " ".join( tokens[start_position:(end_position + 1)]) logging.info("start_position: %d", (start_position)) logging.info("end_position: %d", (end_position)) logging.info("answer: %s", tokenization.printable_text(answer_text)) feature = InputFeatures(unique_id=unique_id, example_index=example_index, doc_span_index=doc_span_index, tokens=tokens, token_to_orig_map=token_to_orig_map, token_is_max_context=token_is_max_context, input_ids=input_ids, input_mask=input_mask, segment_ids=segment_ids, start_position=start_position, end_position=end_position, is_impossible=example.is_impossible) # Run callback if is_training: output_fn(feature) else: output_fn(feature, is_padding=False) unique_id += 1 if not is_training and feature: assert batch_size num_padding = 0 num_examples = unique_id - base_id if unique_id % batch_size != 0: num_padding = batch_size - (num_examples % batch_size) logging.info("Adding padding examples to make sure no partial batch.") logging.info("Adds %d padding examples for inference.", num_padding) dummy_feature = copy.deepcopy(feature) for _ in range(num_padding): dummy_feature.unique_id = unique_id # Run callback output_fn(feature, is_padding=True) unique_id += 1 return unique_id - base_id