viable/strict/1788218631: [MPS] Fix conv backward to preserve input memory format (#174241)
- PyTorch 近 90 天出现 350 次
- PyTorch 近 90 天第 347 次发版
- 上一次:同一天稍早 · viable/strict/1788216615: [inductor] Enable BF16 index_add lowering on SM90+ and ROCm (#195192)
发生了什么
Summary Fixes convolution_backward on MPS to preserve the original input tensor's memory format, rather than deriving it from grad_output . When convolution_backward is called with a channels_last input but contiguous grad_output , the grad_input should preserve the input's memory format ( channels_last ), not adopt grad_output 's format (contiguous). Problem Previously, mps_convolution_backward_input determined the…
摘要按规则整理自下方来源原文