Remove more unnecessary load_weights methods (#47058)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
@@ -709,6 +709,20 @@ class Molmo2VisionBackbone(nn.Module, SupportsQuant):
|
||||
"merged_linear": ["gate_proj", "up_proj"],
|
||||
}
|
||||
|
||||
# Runs after the top-level mapper, so image_pooling_2d/image_projector
|
||||
# source names are already renamed to q/k/v_proj and gate/up_proj.
|
||||
hf_to_vllm_mapper = WeightsMapper(
|
||||
orig_to_new_stacked={
|
||||
"wq": ("merged_qkv", "q"),
|
||||
"wk": ("merged_qkv", "k"),
|
||||
"wv": ("merged_qkv", "v"),
|
||||
"k_proj": ("merged_kv", 0),
|
||||
"v_proj": ("merged_kv", 1),
|
||||
"gate_proj": ("merged_linear", 0),
|
||||
"up_proj": ("merged_linear", 1),
|
||||
},
|
||||
)
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
vit_config: VitConfig,
|
||||
@@ -839,43 +853,8 @@ class Molmo2VisionBackbone(nn.Module, SupportsQuant):
|
||||
]
|
||||
|
||||
def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
|
||||
stacked_params_mapping = [
|
||||
# (param_name, shard_name, shard_id)
|
||||
("merged_qkv", "wq", "q"),
|
||||
("merged_qkv", "wk", "k"),
|
||||
("merged_qkv", "wv", "v"),
|
||||
("merged_kv", "k_proj", 0),
|
||||
("merged_kv", "v_proj", 1),
|
||||
("merged_linear", "gate_proj", 0),
|
||||
("merged_linear", "up_proj", 1),
|
||||
]
|
||||
params_dict = dict(self.named_parameters())
|
||||
loaded_params: set[str] = set()
|
||||
|
||||
for name, loaded_weight in weights:
|
||||
for param_name, weight_name, shard_id in stacked_params_mapping:
|
||||
if weight_name not in name:
|
||||
continue
|
||||
name = name.replace(weight_name, param_name)
|
||||
# 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 = param.weight_loader
|
||||
weight_loader(param, loaded_weight, shard_id)
|
||||
break
|
||||
else:
|
||||
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
|
||||
loader = AutoWeightsLoader(self)
|
||||
return loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
|
||||
|
||||
|
||||
class Molmo2Attention(nn.Module):
|
||||
|
||||
Reference in New Issue
Block a user