frozenleaves/vllm
1
0
Fork 0
mirror of https://github.com/vllm-project/vllm.git synced 2025-10-20 14:53:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Model] use AutoWeightsLoader for gpt2 (#18625)

Browse Source 
Signed-off-by: zt2370 <ztang2370@gmail.com>
This commit is contained in:
ztang2370
2025-05-24 21:36:13 +08:00
committed by GitHub
parent 6d166a8d35
commit 2cd4d58df4
1 changed files with 41 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

69
vllm/model_executor/models/gpt2.py
View File

@ -43,7 +43,7 @@ from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.sequence import IntermediateTensors
from .interfaces import SupportsPP
from .utils import (is_pp_missing_parameter,
from .utils import (AutoWeightsLoader, is_pp_missing_parameter,
make_empty_intermediate_tensors_factory, make_layers,
maybe_prefix)
@ -235,6 +235,35 @@ class GPT2Model(nn.Module):
hidden_states = self.ln_f(hidden_states)
return hidden_states
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
params_dict = dict(self.named_parameters(remove_duplicate=False))
loaded_params: set[str] = set()
for name, loaded_weight in weights:
if ".attn.bias" in name or ".attn.masked_bias" in name:
# Skip attention mask.
# NOTE: "c_attn.bias" should not be skipped.
continue
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
# The HF's GPT-2 implementation uses Conv1D instead of Linear.
# Because of this, we need to transpose the weights.
# Note(zhuohan): the logic below might break quantized models.
for conv1d_weight_name in ["c_attn", "c_proj", "c_fc"]:
if conv1d_weight_name not in name:
continue
if not name.endswith(".weight"):
continue
loaded_weight = loaded_weight.t()
weight_loader = getattr(param, "weight_loader",
default_weight_loader)
weight_loader(param, loaded_weight)
loaded_params.add(name)
return loaded_params
class GPT2LMHeadModel(nn.Module, SupportsPP):
@ -283,32 +312,16 @@ class GPT2LMHeadModel(nn.Module, SupportsPP):
def load_weights(self, weights: Iterable[tuple[str,
torch.Tensor]]) -> set[str]:
params_dict = dict(self.named_parameters(remove_duplicate=False))
loaded_params: set[str] = set()
for name, loaded_weight in weights:
if ".attn.bias" in name or ".attn.masked_bias" in name:
# Skip attention mask.
# NOTE: "c_attn.bias" should not be skipped.
continue
if not name.startswith("transformer.") and not name.startswith(
"lm_head"):
name = "transformer." + name
loader = AutoWeightsLoader(self)
weights = _add_transformer_prefix(weights)
return loader.load_weights(weights)
if is_pp_missing_parameter(name, self):
continue
param = params_dict[name]
# The HF's GPT-2 implementation uses Conv1D instead of Linear.
# Because of this, we need to transpose the weights.
# Note(zhuohan): the logic below might break quantized models.
for conv1d_weight_name in ["c_attn", "c_proj", "c_fc"]:
if conv1d_weight_name not in name:
continue
if not name.endswith(".weight"):
continue
loaded_weight = loaded_weight.t()
weight_loader = getattr(param, "weight_loader",
default_weight_loader)
weight_loader(param, loaded_weight)
loaded_params.add(name)
return loaded_params
def _add_transformer_prefix(
weights: Iterable[tuple[str, torch.Tensor]]
) -> Iterable[tuple[str, torch.Tensor]]:
for name, tensor in weights:
if not name.startswith('transformer.') and not name.startswith(
"lm_head"):
name = 'transformer.' + name
yield name, tensor