llvm-project

02bf3b54 - [mlir][linalg] Add quantized conv2d operator with FCHW,NCHW order (#107740)

[mlir][linalg] Add quantized conv2d operator with FCHW,NCHW order (#107740) This patch adds a quantized version of the `linalg.conv2d_nchw_fchw` Op. This is the "channel-first" ordering typically used by PyTorch and others.