900b9a7f2884aaf419e62508be7497b0ad3e1f62,pytext/metric_reporters/squad_metric_reporter.py,SquadMetricReporter,_unnumberize,#SquadMetricReporter#,358

Before Change

        // start_idx and end_idx are lists of char start and end positions in doc_str.
        doc_tokens, start_idx, end_idx = self.tensorizer._lookup_tokens(doc_str)
        doc_tokens = list(doc_tokens)
        num_ans_tokens = len(ans_tokens)
        start_char_idx = 0
        end_char_idx = 0
        answer_str = ""
        for doc_token_idx in range(len(doc_tokens) - num_ans_tokens):

After Change

        // start_idx and end_idx are lists of char start and end positions in doc_str.
        doc_tokens, start_idxs, end_idxs = self.tensorizer._lookup_tokens(doc_str)
        // find the offset of doc_tokens in tokens
        offset = list(
            map(
                lambda x: tokens[x : x + len(doc_tokens)] == doc_tokens,
                range(len(tokens) - len(doc_tokens) + 1),
            )
        ).index(True)
        assert offset > -1

        // find the answer char idxs
        start_char_idx = 0

Instances