biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Models] Intern-S1-Pro (#33636)

Browse Source 
Signed-off-by: zxy <zhou0493@e.ntu.edu.sg>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
This commit is contained in:
zxy
2026-02-03 21:49:45 +08:00
committed by GitHub
parent be8168ff88
commit a3acfa1071
11 changed files with 942 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
vllm/model_executor/models/qwen3_moe.py
View File

@@ -428,7 +428,13 @@ class Qwen3MoeDecoderLayer(nn.Module):
@support_torch_compile
class Qwen3MoeModel(nn.Module):
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
def __init__(
self,
*,
vllm_config: VllmConfig,
prefix: str = "",
decoder_layer_type: type[torch.nn.Module] = Qwen3MoeDecoderLayer,
):
super().__init__()
config = vllm_config.model_config.hf_text_config
@@ -449,7 +455,7 @@ class Qwen3MoeModel(nn.Module):
)
self.start_layer, self.end_layer, self.layers = make_layers(
config.num_hidden_layers,
lambda prefix: Qwen3MoeDecoderLayer(vllm_config=vllm_config, prefix=prefix),
lambda prefix: decoder_layer_type(vllm_config=vllm_config, prefix=prefix),
prefix=f"{prefix}.layers",
)
self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)