xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-23 19:55:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Mamba] Removed disable cascade attn in MambaModelConfig (#30712)

Browse Source 
Signed-off-by: asafg <39553475+Josephasafg@users.noreply.github.com>
This commit is contained in:
Asaf Joseph Gardin 2025-12-17 10:48:53 +02:00 committed by GitHub
parent 20fda43151
commit a9e15c21ef
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 0 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/model_executor/models/config.py
View File

@ -308,12 +308,6 @@ class MambaModelConfig(VerifyAndUpdateConfig):
if cache_config.mamba_block_size is None: if cache_config.mamba_block_size is None:
cache_config.mamba_block_size = model_config.max_model_len cache_config.mamba_block_size = model_config.max_model_len
# TODO(tdoublep): remove once cascade attention is supported
logger.info(
"Disabling cascade attention since it is not supported for hybrid models."
)
model_config.disable_cascade_attn = True
class HybridAttentionMambaModelConfig(VerifyAndUpdateConfig): class HybridAttentionMambaModelConfig(VerifyAndUpdateConfig):
@classmethod @classmethod