423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2026-11-06 15:10:01 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix(longlora): extend device type check to include Ascend NPU (#10829)

Browse Source
This commit is contained in:
li-lizhe
2026-11-03 20:54:27 +08:00
committed by GitHub
parent 9997aa078b
commit a3e7f60ae6
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
src/llamafactory/model/model_utils/longlora.py
View File

@@ -317,7 +317,7 @@ def llama_sdpa_attention_forward(
if attention_mask is not None: if attention_mask is not None:
causal_mask = causal_mask[:, :, :, : key_states.shape[-2]] causal_mask = causal_mask[:, :, :, : key_states.shape[-2]]
if query_states.device.type == "cuda" and causal_mask is not None: # avoid pytorch bug if query_states.device.type in ("cuda", "npu") and causal_mask is not None: # avoid pytorch bug
query_states = query_states.contiguous() query_states = query_states.contiguous()
key_states = key_states.contiguous() key_states = key_states.contiguous()
value_states = value_states.contiguous() value_states = value_states.contiguous()