[Bugfix] Use latency MOE backend as default for Flashinfer and other misc fixes (#27439)

Signed-off-by: Pavani Majety <pmajety@nvidia.com>
This commit is contained in:
Pavani Majety
2025-11-07 04:18:39 -08:00
committed by GitHub
parent e0919f331d
commit 72b1c2ae2c
7 changed files with 47 additions and 12 deletions
+1 -1
View File
@@ -1385,7 +1385,7 @@ 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.zeros(
output_scale = torch.empty(
(rounded_m, rounded_n // 4), device=device, dtype=torch.int32
)