[Bugfix] Fix gpt-oss w4a8 DP/EP on B200 (#26729)

Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
This commit is contained in:
Varun Sundar Rabindranath
2025-10-21 01:51:14 -04:00
committed by GitHub
co-authored by Varun Sundar Rabindranath Michael Goin
parent f95da13c3d
commit 5ff5d94e77
5 changed files with 82 additions and 2 deletions
@@ -18,4 +18,7 @@ def mxfp8_e4m3_quantize(x: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
"`pip install flashinfer`"
) from err
return mxfp8_e4m3_quantize(x, is_sf_swizzled_layout=False)
x_q, x_scales = mxfp8_e4m3_quantize(x, is_sf_swizzled_layout=False)
if x_scales.ndim == 1:
x_scales = x_scales.view(x.size(0), -1)
return x_q, x_scales