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

Added qwen3 vision language moe support for speculative decoding (#32048)

Browse Source 
Signed-off-by: shanjiaz <zsjwpianpian@gmail.com>
Signed-off-by: shanjiaz <43143795+shanjiaz@users.noreply.github.com>
This commit is contained in:
shanjiaz
2026-01-20 22:24:05 -05:00
committed by GitHub
parent 0900cedb3f
commit 7ab80a8e37
2 changed files with 17 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
vllm/model_executor/models/qwen3_vl_moe.py
View File

@@ -110,9 +110,14 @@ class Qwen3MoeLLMModel(Qwen3MoeModel):
assert intermediate_tensors is not None
hidden_states = intermediate_tensors["hidden_states"]
residual = intermediate_tensors["residual"]
aux_hidden_states = []
for layer_idx, layer in islice(
enumerate(self.layers), self.start_layer, self.end_layer
):
if layer_idx in self.aux_hidden_state_layers:
aux_hidden_states.append(hidden_states + residual)
hidden_states, residual = layer(
positions,
hidden_states,
@@ -132,6 +137,9 @@ class Qwen3MoeLLMModel(Qwen3MoeModel):
{"hidden_states": hidden_states, "residual": residual}
)
hidden_states, _ = self.norm(hidden_states, residual)
if len(aux_hidden_states) > 0:
return hidden_states, aux_hidden_states
return hidden_states
def load_fused_expert_weights(