label_smoothed_encouraging_loss_fairseq.py

# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.

import math

import torch
from fairseq import metrics, utils
from fairseq.criterions import FairseqCriterion, register_criterion
import torch.nn.functional as F
import math # add for encouraging loss


class ARGS():
    def __init__(
            self,
            log_end=0.5,
    ):
        self.log_end=log_end


def label_smoothed_nll_loss(lprobs, target, epsilon, ignore_index=None, reduce=True):
    if target.dim() == lprobs.dim() - 1:
        target = target.unsqueeze(-1)  # 这个是为了gather时用的
    nll_loss = -lprobs.gather(dim=-1, index=target)
    smooth_loss = -lprobs.sum(dim=-1, keepdim=True)
    if ignore_index is not None:
        pad_mask = target.eq(ignore_index)
        nll_loss.masked_fill_(pad_mask, 0.0)
        smooth_loss.masked_fill_(pad_mask, 0.0)
    else:
        nll_loss = nll_loss.squeeze(-1)
        smooth_loss = smooth_loss.squeeze(-1)
    if reduce:
        nll_loss = nll_loss.sum()
        smooth_loss = smooth_loss.sum()
    eps_i = epsilon / lprobs.size(-1)
    # (1-epsilon)*y *-logp +epsilon/N * -logp=((1-epsilon)*y+epsilon/N)*-logp
    #y*log(1-p)->((1-epsilon)*y+epsilon/N)*log(1-p)
    loss = (1.0 - epsilon) * nll_loss + eps_i * smooth_loss
    return loss, nll_loss


@register_criterion("label_smoothed_encourage_loss_fairseq")
class LabelSmoothedEncourageLossFairseq(FairseqCriterion):
    def __init__(
        self,
        task,
        sentence_avg,
        label_smoothing,
        ignore_prefix_size=0,
        report_accuracy=False,
        log_end=1,

    ):
        super().__init__(task)
        self.sentence_avg = sentence_avg
        self.eps = label_smoothing
        self.ignore_prefix_size = ignore_prefix_size
        self.report_accuracy = report_accuracy
        # add for encouraging loss
        self.args = ARGS(log_end=log_end)
        self.set_epoch(1)

    @staticmethod
    def add_args(parser):
        """Add criterion-specific arguments to the parser."""
        # fmt: off
        parser.add_argument('--label-smoothing', default=0., type=float, metavar='D',
                            help='epsilon for label smoothing, 0 means no label smoothing')
        parser.add_argument('--report-accuracy', action='store_true',
                            help='report accuracy metric')
        parser.add_argument('--ignore-prefix-size', default=0, type=int,
                            help='Ignore first N tokens')
        # add for encouraging loss
        parser.add_argument('--log_end', type=float, default=0.5,help=' # 1 refers to the normal bonus, but 0.75 can easily work in existing optimization systems, 0.5 work for all settings we tested, recommend LE=0.75 for high accuracy scenarios and low LE for low accuracy scenarios.')

        # fmt: on

    def forward(self, model, sample, reduce=True):
        """Compute the loss for the given sample.

        Returns a tuple with three elements:
        1) the loss
        2) the sample size, which is used as the denominator for the gradient
        3) logging outputs to display while training
        """
        net_output = model(**sample["net_input"])
        loss, nll_loss = self.compute_loss(model, net_output, sample, reduce=reduce)
        sample_size = (
            sample["target"].size(0) if self.sentence_avg else sample["ntokens"]
        )
        logging_output = {
            "loss": loss.data,
            "nll_loss": nll_loss.data,
            "ntokens": sample["ntokens"],
            "nsentences": sample["target"].size(0),
            "sample_size": sample_size,
        }
        if self.report_accuracy:
            n_correct, total = self.compute_accuracy(model, net_output, sample)
            logging_output["n_correct"] = utils.item(n_correct.data)
            logging_output["total"] = utils.item(total.data)
        return loss, sample_size, logging_output

    def get_lprobs_and_target(self, model, net_output, sample):
        lprobs = model.get_normalized_probs(net_output, log_probs=True)
        target = model.get_targets(sample, net_output)
        if self.ignore_prefix_size > 0:
            if getattr(lprobs, "batch_first", False):
                lprobs = lprobs[:, self.ignore_prefix_size :, :].contiguous()
                target = target[:, self.ignore_prefix_size :].contiguous()
            else:
                lprobs = lprobs[self.ignore_prefix_size :, :, :].contiguous()
                target = target[self.ignore_prefix_size :, :].contiguous()
        return lprobs.view(-1, lprobs.size(-1)), target.view(-1)

    def compute_loss(self, model, net_output, sample, reduce=True):
        lprobs, target = self.get_lprobs_and_target(model, net_output, sample)
        loss, nll_loss = label_smoothed_nll_loss(
            lprobs,
            target,
            self.eps,
            ignore_index=self.padding_idx,
            reduce=reduce,
        )
        # ----- begin for encouraging loss  -----
        probs = torch.exp(lprobs)
        bonus = torch.log(torch.clamp((torch.ones_like(probs) - probs), min=1e-5))  # likelihood bonus
        if self.args.log_end != 1.0:  # e.g. 0.9
            log_end = self.args.log_end
            y_log_end = torch.log(torch.ones_like(probs) - log_end)
            bonus_after_log_end = 1/(log_end - torch.ones_like(probs)) * (probs-log_end) + y_log_end
            bonus = torch.where(probs > log_end, bonus_after_log_end, bonus)
        c_loss = F.nll_loss(
            -bonus,
            target.view(-1),
            reduction='sum',
            ignore_index=self.padding_idx if self.padding_idx is not None else -100
        )  # y*log(1-p)
        smoothing_c_loss = bonus.sum(dim=-1)
        pad_mask = target.view(-1).eq(self.padding_idx)
        smoothing_c_loss.masked_fill_(pad_mask, 0.0)
        smoothing_c_loss = smoothing_c_loss.sum()
        c_loss = c_loss*(1-self.eps) + (self.eps/lprobs.size(-1))*smoothing_c_loss
        loss = loss + c_loss
        # ----end for encouraging loss ----
        return loss, nll_loss

    def compute_accuracy(self, model, net_output, sample):
        lprobs, target = self.get_lprobs_and_target(model, net_output, sample)
        mask = target.ne(self.padding_idx)
        n_correct = torch.sum(
            lprobs.argmax(1).masked_select(mask).eq(target.masked_select(mask))
        )
        total = torch.sum(mask)
        return n_correct, total

    @classmethod
    def reduce_metrics(cls, logging_outputs) -> None:
        """Aggregate logging outputs from data parallel training."""
        loss_sum = sum(log.get("loss", 0) for log in logging_outputs)
        nll_loss_sum = sum(log.get("nll_loss", 0) for log in logging_outputs)
        ntokens = sum(log.get("ntokens", 0) for log in logging_outputs)
        sample_size = sum(log.get("sample_size", 0) for log in logging_outputs)

        metrics.log_scalar(
            "loss", loss_sum / sample_size / math.log(2), sample_size, round=3
        )
        metrics.log_scalar(
            "nll_loss", nll_loss_sum / ntokens / math.log(2), ntokens, round=3
        )
        metrics.log_derived(
            "ppl", lambda meters: utils.get_perplexity(meters["nll_loss"].avg)
        )

        total = utils.item(sum(log.get("total", 0) for log in logging_outputs))
        if total > 0:
            metrics.log_scalar("total", total)
            n_correct = utils.item(
                sum(log.get("n_correct", 0) for log in logging_outputs)
            )
            metrics.log_scalar("n_correct", n_correct)
            metrics.log_derived(
                "accuracy",
                lambda meters: round(
                    meters["n_correct"].sum * 100.0 / meters["total"].sum, 3
                )
                if meters["total"].sum > 0
                else float("nan"),
            )

    @staticmethod
    def logging_outputs_can_be_summed() -> bool:
        """
        Whether the logging outputs returned by `forward` can be summed
        across workers prior to calling `reduce_metrics`. Setting this
        to True will improves distributed training speed.
        """
        return True

    def set_epoch(self,epoch):
        self.epoch = epoch

    def get_epoch(self):
        return self.epoch