dd2eaf6c174fdd757514bc06d1063e5f9c6bb4a0,torchnlp/samplers/bucket_batch_sampler.py,BucketBatchSampler,__iter__,#BucketBatchSampler#,90

Before Change

                    yield batch

        if self.biggest_batches_first is None:
            return get_batches()
        else:
            batches = list(get_batches())
            biggest_batches = heapq.nlargest(
                5,
                range(len(batches)),
                key=lambda i: sum([self.biggest_batches_first(self.data[j]) for j in batches[i]]))
            front = [batches[i] for i in biggest_batches]
            // Remove ``biggest_batches`` from data
            for i in sorted(biggest_batches, reverse=True):
                batches.pop(i)
            // Move them to the front
            batches[0:0] = front
            return iter(batches)

    def __len__(self):
        if self.drop_last:
            return len(self.data) // self.batch_size
        else:

After Change

            sorted_sampler = SortedSampler(bucket, self.sort_key)
            for batch in SubsetRandomSampler(
                    list(BatchSampler(sorted_sampler, self.batch_size, self.drop_last))):
                yield [bucket[i] for i in batch]

    def __len__(self):
        if self.drop_last:
            return len(self.sampler) // self.batch_size

Instances