Umang Yadav

Add logic to unfuse mlir and pointwise module for the split-k GEMM configs

3

Solves #2898

This can be done by shifting zero point by -128.

Convert OCP FP8 model to FNUZ model inside MIGraphX

5

### Problem Description ### Parsing OCP FP8 Model This would require MIGraphX to expose E4M3FN data type into the IR. Currently only E4M3FNUZ type is exposed. It is probably not...

Add TF parsers ResizeByNearestNeighbour

1

These ops appears in EfficientDet model.

`migraphx-driver` doesn't seem to quantize to int8 or fp8 https://github.com/ROCm/AMDMIGraphX/blob/1199fbeeb83539a5cb6a033745e67bfb661875ca/src/driver/verify.cpp#L97 Need to add quantization for 8 bits and combination of `--fp16 --int8`.

FP8 lossy downcast issue with "ref" implementation

3

https://github.com/ROCmSoftwarePlatform/AMDMIGraphX/pull/2506/files This PR had to disable FP8 tests for the CPU backend. Ref implementation is doing Float -- > Fp8 -- > Float conversion but CPU backend is doing entire...

Allow parsing of E4M3FN models using scales manipulation

2

If the E4M3FN model is quantized using QDQ pairs then, it can be converted into E4M3FNUZ types by multiplying scales with some constants. This would allow using E4M3FN models inside...