GitLab

Add int4+scale based on Zhang, Jing pk_i4. Compile pass, function pass.
Modify the kernel to 128x128x128, and use mfma_32x32x4
Move the weight permute from host to device

Modified the scale init method.

Modified the init method, the function is failed, need to debug.

Added init method

Support group=128 for Llam2-7B-int4

Move the weight permute from host to device

Add ckProfiler for GEMM b scale (int4)

Add reference function.

Add pipeline v4 (2 LDS pingpong)

Add more int4-Gemm kernel profiling instances.

Modify the in4-Gemm kernel instances

Move the pk_i4 permute in kernel

* update build logic with GPU_ARCHS

* fix the GPU_ARCHS build for codegen

* unset GPU_TARGETS when GPU_ARCHS are set

Without this change, the following diagnostic is generated:
  a template argument list is expected after a name prefixed by the template
  keyword [-Wmissing-template-arg-list-after-template-kw]

See C++17 spec [temp.names] p5.

* fix clang20 compilation errors for gfx90a

* fix clang20 compilation errors for gfx11 targets

* Support NGCHW in grouped conv fwd

* Remove not needed variable

* Fixes

The dynamic buffer doesn't have support for fp8 in `Update` operation thus fp8 is not supporting `InMemoryDataOperation::Add`