Add CUDA implementation of GatherBlockQuantized operator#25575

Merged

kunal-vaishnavi merged 24 commits intomicrosoft:mainfrom

xiaomsft:xiaoh/gather_block_quantized_cuda

Aug 1, 2025

Commits on Aug 1, 2025

Add CUDA implementation of GatherBlockQuantized operator
Xiaoyan Hu
committed
Clang format
Xiaoyan Hu
committed
Run with lint runner
Xiaoyan Hu
committed
Add indices shape check
Xiaoyan Hu
committed
Fix webgpu tests
Xiaoyan Hu
committed
Disable webgpu tests which are failing
Xiaoyan Hu
committed
remove unused formal parameter
Xiaoyan Hu
committed
void unused parameter
Xiaoyan Hu
committed
void unused parameter
Xiaoyan Hu
committed
Fix nodiscard status
Xiaoyan Hu
committed
Fix integer precision
Xiaoyan Hu
committed
cast raw data instead of using template for data fetching
Xiaoyan Hu
committed
Fix CUDA_MS_OP_THREE_TYPED_CLASS_NAME
Xiaoyan Hu
committed
Fix mismatched marcro name
Xiaoyan Hu
committed
Fix BuildKernelCreateInfo
Xiaoyan Hu
committed
Use raw data
Xiaoyan Hu
committed
Fix typo
Xiaoyan Hu
committed
Fix axis check
Xiaoyan Hu
committed
Fix kernel
Xiaoyan Hu
committed
Support both signed and unsigned data type
Xiaoyan Hu
committed
Fix templates
Xiaoyan Hu
committed
Fix sign type
Xiaoyan Hu
committed
Disable more tests
Xiaoyan Hu
committed
Update autogenerated doc
Xiaoyan Hu
committed