Python sequence_scorer.SequenceScorer方法代码示例

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def build_generator(self, models, args):
        if getattr(args, 'score_reference', False):
            from fairseq.sequence_scorer import SequenceScorer
            return SequenceScorer(
                self.target_dictionary,
                eos=self.tgt_dict.index('[{}]'.format(self.args.target_lang))
            )
        else:
            from fairseq.sequence_generator import SequenceGenerator
            return SequenceGenerator(
                models,
                self.target_dictionary,
                beam_size=getattr(args, 'beam', 5),
                max_len_a=getattr(args, 'max_len_a', 0),
                max_len_b=getattr(args, 'max_len_b', 200),
                min_len=getattr(args, 'min_len', 1),
                normalize_scores=(not getattr(args, 'unnormalized', False)),
                len_penalty=getattr(args, 'lenpen', 1),
                unk_penalty=getattr(args, 'unkpen', 0),
                temperature=getattr(args, 'temperature', 1.),
                match_source_len=getattr(args, 'match_source_len', False),
                no_repeat_ngram_size=getattr(args, 'no_repeat_ngram_size', 0),
                eos=self.tgt_dict.index('[{}]'.format(self.args.target_lang))
            ) 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def build_generator(self, models, args):
        if getattr(args, 'score_reference', False):
            from fairseq.sequence_scorer import SequenceScorer
            return SequenceScorer(
                self.target_dictionary,
                eos=self.tgt_dict.index('[{}]'.format(self.target_lang))
            )
        else:
            from fairseq.sequence_generator import SequenceGenerator
            return SequenceGenerator(
                models,
                self.target_dictionary,
                beam_size=getattr(args, 'beam', 5),
                max_len_a=getattr(args, 'max_len_a', 0),
                max_len_b=getattr(args, 'max_len_b', 200),
                min_len=getattr(args, 'min_len', 1),
                normalize_scores=(not getattr(args, 'unnormalized', False)),
                len_penalty=getattr(args, 'lenpen', 1),
                unk_penalty=getattr(args, 'unkpen', 0),
                temperature=getattr(args, 'temperature', 1.),
                match_source_len=getattr(args, 'match_source_len', False),
                no_repeat_ngram_size=getattr(args, 'no_repeat_ngram_size', 0),
                eos=self.tgt_dict.index('[{}]'.format(self.args.target_lang))
            ) 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def __init__(self, path, data, use_cpu=True):
        # Create the language modeling task.
        self.args = FluencyArgs(path, data)
        self.task = tasks.setup_task(self.args)
        self.use_cuda = torch.cuda.is_available and not use_cpu

        # Load language model ensemble.
        models, model_args = utils.load_ensemble_for_inference(self.args.path.split(':'), self.task)
        self.models = models
        self.model_args = model_args

        # Optimize ensemble for generation.
        for model in self.models:
            model.make_generation_fast_()
            if self.use_cuda and self.model_args.fp16:
                model.half()
        
        # Create the sequence scorer.
        self.scorer = SequenceScorer(self.models, self.task.target_dictionary)
        if self.use_cuda:
            self.scorer.cuda() 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def build_generator(self, args):
        """Replace SequenceGenerator with SequenceCopyGenerator."""
        if args.score_reference:
            from fairseq.sequence_scorer import SequenceScorer
            return SequenceScorer(self.target_dictionary)
        else:
            from fairseq.sequence_copygenerator import SequenceCopyGenerator
            return SequenceCopyGenerator(
                self.target_dictionary,
                beam_size=args.beam,
                max_len_a=args.max_len_a,
                max_len_b=args.max_len_b,
                min_len=args.min_len,
                stop_early=(not args.no_early_stop),
                normalize_scores=(not args.unnormalized),
                len_penalty=args.lenpen,
                unk_penalty=args.unkpen,
                sampling=args.sampling,
                sampling_topk=args.sampling_topk,
                sampling_temperature=args.sampling_temperature,
                diverse_beam_groups=args.diverse_beam_groups,
                diverse_beam_strength=args.diverse_beam_strength,
                match_source_len=args.match_source_len,
                no_repeat_ngram_size=args.no_repeat_ngram_size,
            ) 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def build_generator(self, args):
        if args.score_reference:
            from fairseq.sequence_scorer import SequenceScorer
            return SequenceScorer(self.target_dictionary)
        else:
            from fairseq.sequence_generator import SequenceGenerator
            return SequenceGenerator(
                self.target_dictionary,
                beam_size=args.beam,
                max_len_a=args.max_len_a,
                max_len_b=args.max_len_b,
                min_len=args.min_len,
                stop_early=(not args.no_early_stop),
                normalize_scores=(not args.unnormalized),
                len_penalty=args.lenpen,
                unk_penalty=args.unkpen,
                sampling=args.sampling,
                sampling_topk=args.sampling_topk,
                sampling_temperature=args.sampling_temperature,
                diverse_beam_groups=args.diverse_beam_groups,
                diverse_beam_strength=args.diverse_beam_strength,
                match_source_len=args.match_source_len,
                no_repeat_ngram_size=args.no_repeat_ngram_size,
            ) 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def __init__(self, parsed_args):
        self.args = parsed_args
        import_user_module(parsed_args)
        assert parsed_args.path is not None, '--path required for evaluation'

        print(parsed_args)

        self.use_cuda = torch.cuda.is_available() and not parsed_args.cpu

        self.task = tasks.setup_task(parsed_args)

        # Load ensemble
        print('| loading model(s) from {}'.format(parsed_args.path))
        self.models, args = utils.load_ensemble_for_inference(
            parsed_args.path.split(':'), self.task, model_arg_overrides=eval(parsed_args.model_overrides),
        )

        for model in self.models:
            model.make_generation_fast_()
            if self.use_cuda:
                model.cuda()

        for arg in vars(parsed_args).keys():
            if arg not in {'self_target', 'future_target', 'past_target', 'tokens_per_sample',
                           'output_size_dictionary'}:
                setattr(args, arg, getattr(parsed_args, arg))
        self.task = tasks.setup_task(args)

        self.gen_timer = StopwatchMeter()
        self.scorer = SequenceScorer(self.task.target_dictionary) 

# 需要导入模块: from fairseq import sequence_scorer [as 别名]
# 或者: from fairseq.sequence_scorer import SequenceScorer [as 别名]
def main(args):
    assert args.path is not None, '--path required for evaluation!'

    if args.tokens_per_sample is None:
        args.tokens_per_sample = 1024
    print(args)

    use_cuda = torch.cuda.is_available() and not args.cpu

    # Load dataset splits
    task = tasks.setup_task(args)
    task.load_dataset(args.gen_subset)
    print('| {} {} {} examples'.format(args.data, args.gen_subset, len(task.dataset(args.gen_subset))))

    # Load ensemble
    print('| loading model(s) from {}'.format(args.path))
    models, _ = utils.load_ensemble_for_inference(args.path.split(':'), task)

    # Optimize ensemble for generation and set the source and dest dicts on the model (required by scorer)
    for model in models:
        model.make_generation_fast_()

    itr = data.EpochBatchIterator(
        dataset=task.dataset(args.gen_subset),
        max_sentences=args.max_sentences or 4,
        max_positions=model.max_positions(),
        num_shards=args.num_shards,
        shard_id=args.shard_id,
    ).next_epoch_itr(shuffle=False)

    gen_timer = StopwatchMeter()
    scorer = SequenceScorer(models, task.target_dictionary)
    if use_cuda:
        scorer.cuda()

    score_sum = 0.
    count = 0
    with progress_bar.build_progress_bar(args, itr) as t:
        results = scorer.score_batched_itr(t, cuda=use_cuda, timer=gen_timer)
        wps_meter = TimeMeter()
        for _, src_tokens, __, hypos in results:
            for hypo in hypos:
                pos_scores = hypo['positional_scores']
                inf_scores = pos_scores.eq(float('inf')) | pos_scores.eq(float('-inf'))
                if inf_scores.any():
                    print('| Skipping tokens with inf scores:',
                          task.target_dictionary.string(hypo['tokens'][inf_scores.nonzero()]))
                    pos_scores = pos_scores[(~inf_scores).nonzero()]
                score_sum += pos_scores.sum()
                count += pos_scores.numel()
            wps_meter.update(src_tokens.size(0))
            t.log({'wps': round(wps_meter.avg)})

    avg_nll_loss = -score_sum / count
    print('| Evaluated {} tokens in {:.1f}s ({:.2f} tokens/s)'.format(gen_timer.n, gen_timer.sum, 1. / gen_timer.avg))
    print('| Loss: {:.4f}, Perplexity: {:.2f}'.format(avg_nll_loss, np.exp(avg_nll_loss)))