[LoRA] Optimize 3D MoE logic (#29222)

Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
This commit is contained in:
Jee Jee Li
2025-11-24 10:27:23 +08:00
committed by GitHub
parent c309bb5245
commit 1073ba68b0
11 changed files with 397 additions and 105 deletions
+1
View File
@@ -656,6 +656,7 @@ class GptOssModel(nn.Module):
class GptOssForCausalLM(nn.Module, SupportsPP, SupportsEagle3, SupportsLoRA):
is_3d_moe_weight: bool = True
packed_modules_mapping = {"qkv_proj": ["q_proj", "k_proj", "v_proj"]}
hf_to_vllm_mapper = WeightsMapper(