[V1]SupportsV0Only protocol for model definitions (#13959)

Signed-off-by: Roger Wang <ywang@roblox.com>
This commit is contained in:
Roger Wang
2025-02-27 20:02:15 -05:00
committed by GitHub
parent 67fc426845
commit 6c85da3a18
19 changed files with 93 additions and 32 deletions
+2 -1
View File
@@ -43,6 +43,7 @@ from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsV0Only
from .utils import maybe_prefix
logger = logging.get_logger(__name__)
@@ -776,7 +777,7 @@ class BartModel(nn.Module):
return decoder_outputs
class BartForConditionalGeneration(nn.Module):
class BartForConditionalGeneration(nn.Module, SupportsV0Only):
base_model_prefix = "model"
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):