fairseq/criterions/adaptive_loss.py

Survived

Mutant 2675

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -11,7 +11,7 @@
 from fairseq.criterions import FairseqCriterion, register_criterion
 
 
-@register_criterion('adaptive_loss')
+@register_criterion('XXadaptive_lossXX')
 class AdaptiveLoss(FairseqCriterion):
     """This is an implementation of the loss function accompanying the adaptive softmax approximation for
     graphical processing units (GPU), described in the paper "Efficient softmax approximation for GPUs"

Mutant 2676

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -10,8 +10,6 @@
 from fairseq import metrics, utils
 from fairseq.criterions import FairseqCriterion, register_criterion
 
-
-@register_criterion('adaptive_loss')
 class AdaptiveLoss(FairseqCriterion):
     """This is an implementation of the loss function accompanying the adaptive softmax approximation for
     graphical processing units (GPU), described in the paper "Efficient softmax approximation for GPUs"

Mutant 2677

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -21,7 +21,6 @@
         super().__init__(task)
         self.sentence_avg = sentence_avg
 
-    @classmethod
     def build_criterion(cls, args, task):
         if getattr(args, 'ddp_backend', None) == 'c10d':
             raise Exception(

Mutant 2678

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -31,7 +31,7 @@
             )
         return cls(task, args.sentence_avg)
 
-    def forward(self, model, sample, reduce=True):
+    def forward(self, model, sample, reduce=False):
         """Compute the loss for the given sample.
 
         Returns a tuple with three elements:

Mutant 2679

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -77,7 +77,6 @@
         }
         return loss, sample_size, logging_output
 
-    @staticmethod
     def reduce_metrics(logging_outputs) -> None:
         """Aggregate logging outputs from data parallel training."""
         loss_sum = utils.item(sum(log.get('loss', 0) for log in logging_outputs))

Mutant 2680

--- fairseq/criterions/adaptive_loss.py
+++ fairseq/criterions/adaptive_loss.py
@@ -91,7 +91,6 @@
         else:
             metrics.log_derived('ppl', lambda meters: utils.get_perplexity(meters['loss'].avg))
 
-    @staticmethod
     def logging_outputs_can_be_summed() -> bool:
         """
         Whether the logging outputs returned by `forward` can be summed