Remove unnecessary load_weights methods (#44589)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2026-06-29 01:52:23 -07:00
committed by GitHub
parent db28ae2d07
commit 5051698e41
54 changed files with 821 additions and 1975 deletions
+4 -22
View File
@@ -25,7 +25,6 @@ from vllm.model_executor.layers.vocab_parallel_embedding import (
ParallelLMHead,
VocabParallelEmbedding,
)
from vllm.model_executor.model_loader.weight_utils import default_weight_loader
from vllm.model_executor.models.interfaces import (
HasInnerState,
IsAttentionFree,
@@ -35,7 +34,7 @@ from vllm.sequence import IntermediateTensors
from .utils import (
AutoWeightsLoader,
is_pp_missing_parameter,
WeightsMapper,
make_empty_intermediate_tensors_factory,
make_layers,
maybe_prefix,
@@ -167,29 +166,12 @@ class Mamba2Model(nn.Module):
return hidden_states
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
params_dict = dict(self.named_parameters())
loaded_params: set[str] = set()
for name, loaded_weight in weights:
if "A_log" in name:
name = name.replace("A_log", "A")
# Skip loading extra bias for GPTQ models.
if name.endswith(".bias") and name not in params_dict:
continue
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
weight_loader = getattr(param, "weight_loader", default_weight_loader)
weight_loader(param, loaded_weight)
loaded_params.add(name)
return loaded_params
class Mamba2ForCausalLM(
nn.Module, HasInnerState, IsAttentionFree, SupportsMambaPrefixCaching
):
hf_to_vllm_mapper = WeightsMapper(orig_to_new_substr={".A_log": ".A"})
@classmethod
def get_mamba_state_dtype_from_config(
cls,
@@ -292,4 +274,4 @@ class Mamba2ForCausalLM(
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
loader = AutoWeightsLoader(self)
return loader.load_weights(weights)
return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)