Repository navigation

[WebGPU] Support block-quantized FP8 dense MatMul - #33074

Merged

kunal-vaishnavi merged 13 commits into

mainfrom

sayanshaw/webgpu-fp8-block-matmul

Oct 19, 2026

Merged

[WebGPU] Support block-quantized FP8 dense MatMul#33074
kunal-vaishnavi merged 13 commits into
mainfrom
sayanshaw/webgpu-fp8-block-matmul

Commits on Oct 2, 2026
- Add WebGPU block-quantized FP8 dense matmul
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  cbcd5e9
- Preserve FP16-only ops and select registered WebGPU plugin in Python
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  255d4b3
Commits on Oct 5, 2026
- Make FP8 activation rounding portable across WebGPU backends
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  3e99036
- Declare FP8 activation shader type dependencies
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  c6ca2e2
- Merge remote-tracking branch 'origin/main' into webgpu-fp8-block-matmul
  Sayan Shaw (sayanshaw24)
  committed
  b7c0e85
- Compare FP8 activation ties in input space
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  2f12bcd
Commits on Oct 6, 2026
Commits on Oct 7, 2026
- Merge main with ONNX materialized-node test fix
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  f86ef3c
Commits on Oct 8, 2026
- Merge remote-tracking branch 'origin/main' into webgpu-fp8-block-matmul
  Sayan Shaw (sayanshaw24)
  committed
  9268738
Commits on Oct 16, 2026
- Merge current main into WebGPU block-FP8 dense MatMul
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  d9718cf
- Preserve FP8 activation NaNs and gate subgroup ID shader inputs
  
  Sayan Shaw (sayanshaw24)
  and
  Copilot
  committed
  4905cd0