# Copyright (c) DP Technology.
# 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
import torch.nn.functional as F
from unicore import metrics
from unicore.losses import UnicoreLoss, register_loss

@register_loss("cross_entropy")
class CrossEntropyLoss(UnicoreLoss):
    def __init__(self, task):
        super().__init__(task)

    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 = self.compute_loss(model, net_output, sample, reduce=reduce)
        sample_size = sample["target"].size(0)
        logging_output = {
            "loss": loss.data,
            "bsz": sample["target"].size(0),
            "sample_size": sample_size,
        }
        return loss, sample_size, logging_output

    def compute_loss(self, model, net_output, sample, reduce=True):
        lprobs = F.log_softmax(net_output.float(), dim=-1)
        lprobs = lprobs.view(-1, lprobs.size(-1))
        target = sample['target'].view(-1)
        loss = F.nll_loss(
            lprobs,
            target,
            reduction="sum" if reduce else "none",
        )
        return loss

    @staticmethod
    def reduce_metrics(logging_outputs, split='valid') -> None:
        """Aggregate logging outputs from data parallel training."""
        loss_sum = sum(log.get("loss", 0) for log in logging_outputs)
        sample_size = sum(log.get("sample_size", 0) for log in logging_outputs)

        # we divide by log(2) to convert the loss from base e to base 2
        metrics.log_scalar(
            "loss", loss_sum / sample_size / math.log(2), sample_size, round=3
        )

    @staticmethod
    def logging_outputs_can_be_summed(is_train) -> 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