[Bugfix] Enable padded FP4 quantization (#25947)

Signed-off-by: Roi Koren <roik@nvidia.com>
This commit is contained in:
roikoren755
2025-10-09 10:59:41 -07:00
committed by GitHub
parent 0d37450eb7
commit 4069db3f2e
2 changed files with 1 additions and 3 deletions
+1 -1
View File
@@ -1384,7 +1384,7 @@ def scaled_fp4_quant(
rounded_m = round_up(m, 128)
scale_n = n // block_size
rounded_n = round_up(scale_n, 4)
output_scale = torch.empty(
output_scale = torch.zeros(
(rounded_m, rounded_n // 4), device=device, dtype=torch.int32
)