Skip to content

GitLab

Unverified Commit 6dd5e528 authored by Kyle Sayers's avatar Kyle Sayers Committed by GitHub
Browse files

Squelch MLA warning for Compressed-Tensors Models (#12704) 


Signed-off-by: default avatarKyle Sayers <kylesayrs@gmail.com>
parent c11de33d
Hide whitespace changes
Inline Side-by-side
Showing with 4 additions and 2 deletions
vllm/config.py
View file @ 6dd5e528
...@@ -986,6 +986,9 @@ class ModelConfig: ...@@ -986,6 +986,9 @@ class ModelConfig:
@property @property
def use_mla(self) -> bool: def use_mla(self) -> bool:
if not self.is_deepseek_mla or envs.VLLM_MLA_DISABLE:
return False
if self.quantization is not None and self.quantization not in [\ if self.quantization is not None and self.quantization not in [\
"fp8", "compressed-tensors"]: "fp8", "compressed-tensors"]:
logger.warning( logger.warning(
...@@ -1012,8 +1015,7 @@ class ModelConfig: ...@@ -1012,8 +1015,7 @@ class ModelConfig:
quant_config) quant_config)
return False return False
use_mla = (self.is_deepseek_mla and not envs.VLLM_MLA_DISABLE) return True
return use_mla
@property @property
def supported_runner_types(self) -> Set[RunnerType]: def supported_runner_types(self) -> Set[RunnerType]:
......
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