A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 06:30 UTC
[BugFix] fix FusedMoE.make_expert_params_mapping in EXAONE-MoE (#32196)
Signed-off-by: lkm2835 <lkm2835@gmail.com>
Kyungmin Lee committed
Jan 13, 2026 at 03:00 UTC
3d962d72ab01b7eb210a4719c24f1dd98d5c5c34
1 file changed
+1
vllm/model_executor/models/exaone_moe.py
+1
@@ -338,6 +338,7 @@ class ExaoneMoeModel(nn.Module):
338
# Params for weights, fp8 weight scales, fp8 activation scales
339
# (param_name, weight_name, expert_id, shard_id)
340
return FusedMoE.make_expert_params_mapping(
341
+ self,
342
ckpt_gate_proj_name="gate_proj",
343
ckpt_down_proj_name="down_proj",
344
ckpt_up_proj_name="up_proj",