[MM][Bugfix] Add error log for VL models when enabling FLASHCOMM (#4222)
### What this PR does / why we need it? Add error log for VL models when enabling `VLLM_ASCEND_ENABLE_FLASHCOMM1=1` or `VLLM_ASCEND_ENABLE_FLASHCOMM=1` (for backward compatibility). This is a temporary fix for https://github.com/vllm-project/vllm-ascend/issues/4132. ### Does this PR introduce _any_ user-facing change? ### How was this patch tested? Signed-off-by: shen-shanshan <467638484@qq.com>
This commit is contained in:
@@ -32,7 +32,7 @@ from vllm_ascend.ascend_config import (check_ascend_config, get_ascend_config,
|
||||
from vllm_ascend.torchair.utils import (check_torchair_cache_exist,
|
||||
delete_torchair_cache_file)
|
||||
from vllm_ascend.utils import (ASCEND_QUANTIZATION_METHOD, enable_sp, is_310p,
|
||||
update_aclgraph_sizes,
|
||||
is_vl_model, update_aclgraph_sizes,
|
||||
update_default_aclgraph_sizes)
|
||||
|
||||
if TYPE_CHECKING:
|
||||
@@ -303,6 +303,14 @@ class NPUPlatform(Platform):
|
||||
vllm_config.scheduler_config)
|
||||
vllm_config.scheduler_config = recompute_scheduler_config
|
||||
|
||||
if is_vl_model(vllm_config):
|
||||
if bool(int(os.getenv("VLLM_ASCEND_ENABLE_FLASHCOMM", '0'))) or \
|
||||
bool(int(os.getenv("VLLM_ASCEND_ENABLE_FLASHCOMM1", '0'))):
|
||||
raise ValueError(
|
||||
"Currently, VL models doesn't support "
|
||||
"FLASHCOMM in vllm-ascend. We will fix this in the future. "
|
||||
"Please set VLLM_ASCEND_ENABLE_FLASHCOMM1=0.")
|
||||
|
||||
@classmethod
|
||||
def get_attn_backend_cls(
|
||||
cls,
|
||||
|
||||
Reference in New Issue
Block a user