ZIBEE/bi_150-vllm
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Upgrade to vllm 0.17.0 corex v4.1 overlay

Browse Source
This commit is contained in:
LZiBee
2026-04-29 19:38:22 +08:00
parent 8fac6062e4
commit 938d0854a5
430 changed files with 35969 additions and 14511 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
vllm/model_executor/parameter.py
View File

@@ -8,10 +8,7 @@ from weakref import WeakValueDictionary
import torch
from torch.nn import Parameter
from vllm.distributed import (
get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
)
from vllm.distributed import get_tensor_model_parallel_rank, get_tensor_model_parallel_world_size
from vllm.logger import init_logger
__all__ = [
@@ -197,7 +194,7 @@ class _ColumnvLLMParameter(BasevLLMParameter):
self.output_dim, shard_id * shard_size, shard_size
)
assert param_data.shape == loaded_weight.shape
assert param_data.shape == loaded_weight.shape, f"param_data.shape({param_data.shape}) != loaded_weight.shape({loaded_weight.shape})"
param_data.copy_(loaded_weight)
@@ -218,16 +215,24 @@ class RowvLLMParameter(BasevLLMParameter):
return self._input_dim
def load_row_parallel_weight(self, loaded_weight: torch.Tensor):
shard_size = self.data.shape[self.input_dim]
loaded_weight = loaded_weight.narrow(
self.input_dim, self.tp_rank * shard_size, shard_size
)
tp_rank = get_tensor_model_parallel_rank()
if len(loaded_weight.shape) == 0:
k = (loaded_weight.shape[0] if loaded_weight.ndim == 1
else loaded_weight.shape[self.input_dim])
assert k % self.tp_size == 0, (
f"Row dimension({k}) must be divisible by tp_size({self.tp_size})!")
shard_size = k // self.tp_size
start_idx = tp_rank * shard_size
loaded_weight = loaded_weight.narrow(self.input_dim, start_idx, shard_size)
if loaded_weight.ndim == 0:
loaded_weight = loaded_weight.reshape(1)
assert self.data.shape == loaded_weight.shape
self.data.copy_(loaded_weight)
if self.data.shape == loaded_weight.shape:
self.data.copy_(loaded_weight)
else:
target_slice = self.data.narrow(self.input_dim, 0, shard_size)
target_slice.copy_(loaded_weight)
class ModelWeightParameter(_ColumnvLLMParameter, RowvLLMParameter):