viable/strict/1788218631: [MPS] Fix conv backward to preserve input memory format (#174241)
- PyTorch: 344 events in the last 90 days
- PyTorch: 341th Release in the last 90 days
- Previous: earlier the same day · viable/strict/1788216615: [inductor] Enable BF16 index_add lowering on SM90+ and ROCm (#195192)
What happened
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…
Summary assembled by rule from the sources below