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

[V1][Hybrid] Mamba Prefix Caching with align mode (#30877)

Browse Source 
Signed-off-by: huanghaoyan.hhy <huanghaoyan.hhy@alibaba-inc.com>
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>
This commit is contained in:
Harry Huang
2026-01-24 01:56:48 +08:00
committed by GitHub
parent fec9da0af4
commit 5206e5e28c
42 changed files with 1774 additions and 128 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/model_executor/models/nano_nemotron_vl.py
View File

@@ -2128,3 +2128,7 @@ class NemotronH_Nano_VL_V2(
temp_vllm_config = copy.deepcopy(vllm_config)
temp_vllm_config.model_config.hf_config = text_config
return NemotronHForCausalLM.get_mamba_state_dtype_from_config(temp_vllm_config)
@classmethod
def get_mamba_state_copy_func(cls):
return NemotronHForCausalLM.get_mamba_state_copy_func()