xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-04-09 10:47:06 +08:00
Code Issues Packages Projects Releases Wiki Activity

updated

Browse Source 
Signed-off-by: Robert Shaw <robshaw@redhat.com>
This commit is contained in:
Robert Shaw 2025-12-24 17:48:41 +00:00
parent 5d93089686
commit 3de8a858b3
1 changed files with 4 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
vllm/model_executor/layers/fused_moe/rocm_aiter_fused_moe.py
View File

@ -325,8 +325,11 @@ class AiterExperts(mk.FusedMoEPermuteExpertsUnpermute):
expert_tokens_meta: mk.ExpertTokensMetadata | None,
apply_router_weight_on_input: bool,
):
# TODO(rob): rocm_aiter_fused_experts uses self.quant_config's
# a_scales for static quantization. Update this to fit better
# with the interface once all quant integrations are complete.
assert a1q_scale is None
assert a2_scale is None
assert a2_scale == self.quant_config.a2_scale
assert expert_tokens_meta is None
result = rocm_aiter_fused_experts(