xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-31 03:07:04 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Bugfix] Fix ColumnParallelLinearWithLoRA slice (#11708)

Browse Source 
Signed-off-by: ZincCat <zincchloride@outlook.com>
This commit is contained in:
ZincCat 2025-01-03 13:02:34 -08:00 committed by GitHub
parent 80c751e7f6
commit 61fed92c7e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
vllm/lora/layers.py
View File

@ -479,7 +479,7 @@ class ColumnParallelLinearWithLoRA(BaseLinearLayerWithLoRA):
# ColumnParallelLinear. # ColumnParallelLinear.
else: else:
tensor_model_parallel_rank = get_tensor_model_parallel_rank() tensor_model_parallel_rank = get_tensor_model_parallel_rank()
shard_size = self.output_dim shard_size = self.output_size
start_idx = tensor_model_parallel_rank * shard_size start_idx = tensor_model_parallel_rank * shard_size
end_idx = (tensor_model_parallel_rank + 1) * shard_size end_idx = (tensor_model_parallel_rank + 1) * shard_size
lora_b = lora_b[:, start_idx:end_idx] lora_b = lora_b[:, start_idx:end_idx]
@ -490,7 +490,7 @@ class ColumnParallelLinearWithLoRA(BaseLinearLayerWithLoRA):
if bias is None: if bias is None:
return bias return bias
tensor_model_parallel_rank = get_tensor_model_parallel_rank() tensor_model_parallel_rank = get_tensor_model_parallel_rank()
shard_size = self.output_dim shard_size = self.output_size
start_idx = tensor_model_parallel_rank * shard_size start_idx = tensor_model_parallel_rank * shard_size
end_idx = (tensor_model_parallel_rank + 1) * shard_size end_idx = (tensor_model_parallel_rank + 1) * shard_size
bias = bias[start_idx:end_idx] bias = bias[start_idx:end_idx]