Reduce AutoWeightsLoader kwargs (#53106)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -41,7 +41,7 @@ from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
|
||||
from vllm.model_executor.models.qwen2 import Qwen2ForCausalLM, Qwen2Model
|
||||
from vllm.sequence import IntermediateTensors
|
||||
|
||||
from .utils import AutoWeightsLoader, PPMissingLayer, maybe_prefix
|
||||
from .utils import AutoWeightsLoader, PPMissingLayer, WeightsMapper, maybe_prefix
|
||||
|
||||
logger = init_logger(__name__)
|
||||
|
||||
@@ -87,6 +87,9 @@ class MiMoModel(Qwen2Model):
|
||||
|
||||
|
||||
class MiMoForCausalLM(Qwen2ForCausalLM, nn.Module):
|
||||
# MTP layers are loaded by the draft model, not the main model.
|
||||
hf_to_vllm_mapper = WeightsMapper(orig_to_new_prefix={"model.mtp_layers.": None})
|
||||
|
||||
def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
|
||||
nn.Module.__init__(self)
|
||||
config = vllm_config.model_config.hf_config
|
||||
@@ -119,9 +122,8 @@ class MiMoForCausalLM(Qwen2ForCausalLM, nn.Module):
|
||||
)
|
||||
|
||||
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
|
||||
# MTP layers are loaded by the draft model, not the main model.
|
||||
loader = AutoWeightsLoader(self, skip_prefixes=["model.mtp_layers."])
|
||||
return loader.load_weights(weights)
|
||||
loader = AutoWeightsLoader(self)
|
||||
return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
|
||||
|
||||
def compute_logits(
|
||||
self,
|
||||
|
||||
Reference in New Issue
Block a user