Transformers backend already supports V1 (#15463)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -1613,14 +1613,6 @@ class EngineArgs:
|
||||
recommend_to_remove=False)
|
||||
return False
|
||||
|
||||
# No TransformersModel support so far.
|
||||
if (model_config.model_impl == ModelImpl.TRANSFORMERS
|
||||
or model_config.model_impl == "transformers"):
|
||||
_raise_or_fallback(
|
||||
feature_name=f"model_impl={model_config.model_impl}",
|
||||
recommend_to_remove=False)
|
||||
return False
|
||||
|
||||
# No Concurrent Partial Prefills so far.
|
||||
if (self.max_num_partial_prefills
|
||||
!= EngineArgs.max_num_partial_prefills
|
||||
|
||||
@@ -24,6 +24,7 @@ from transformers import AutoModel, PretrainedConfig, PreTrainedModel
|
||||
from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS
|
||||
|
||||
from vllm.attention import Attention
|
||||
from vllm.compilation.decorators import support_torch_compile
|
||||
from vllm.config import (CacheConfig, DeviceConfig, ModelConfig,
|
||||
ParallelConfig, VllmConfig)
|
||||
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
|
||||
@@ -109,6 +110,7 @@ def replace_linear_class(
|
||||
)
|
||||
|
||||
|
||||
@support_torch_compile
|
||||
class TransformersModel(nn.Module, SupportsQuant, SupportsLoRA, SupportsPP):
|
||||
embedding_padding_modules = ["lm_head"]
|
||||
embedding_modules = ["embed_tokens"
|
||||
|
||||
Reference in New Issue
Block a user