LinearCrossEntropyLoss

PyTorch 2.13, released July 8, brings FlexAttention to Apple Silicon with up to 12x speedup on sparse attention patterns, such as a 32,768-token sequence with a 256-token sliding window (35ms vs 431ms…

PyTorch 2.13, released this week, brings FlexAttention support for Apple Silicon with up to 12x speedup on sparse patterns and a new fused LinearCrossEntropyLoss operator that cuts peak GPU memory by …