[SM100] Enable fp8 compute for prefill MLA (#30746)

Signed-off-by: Pavani Majety <pmajety@nvidia.com>
This commit is contained in:
Pavani Majety
2025-12-22 19:15:57 +00:00
committed by GitHub
parent 7b926e8901
commit b10f41c894
3 changed files with 117 additions and 18 deletions
@@ -325,7 +325,6 @@ def flashinfer_trtllm_fp4_moe(
local_expert_offset=layer.ep_rank * layer.local_num_experts,
local_num_experts=layer.local_num_experts,
routed_scaling_factor=None,
tile_tokens_dim=None,
routing_method_type=routing_method_type,
do_finalize=True,
)[0]