Skip to content

GitLab

Unverified Commit 9615575a authored by Roger Wang's avatar Roger Wang Committed by GitHub
Browse files

[Bugfix] Fix mamba cache dtype for Qwen3.5 (#34200) 


Signed-off-by: default avatarRoger Wang <hey@rogerw.io>
parent 4293c00b
Hide whitespace changes
Inline Side-by-side
Showing with 2 additions and 1 deletion
vllm/model_executor/models/qwen3_5.py
View file @ 9615575a
......@@ -870,8 +870,9 @@ class Qwen3_5ForConditionalGeneration(Qwen3VLForConditionalGeneration, IsHybrid)
cls,
vllm_config: "VllmConfig",
) -> tuple[torch.dtype, torch.dtype]:
mamba_ssm_dtype = vllm_config.model_config.hf_text_config.mamba_ssm_dtype
return MambaStateDtypeCalculator.gated_delta_net_state_dtype(
vllm_config.model_config.dtype, vllm_config.cache_config.mamba_cache_dtype
vllm_config.model_config.dtype, mamba_ssm_dtype
)
@classmethod
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment