A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 21:00 UTC
[BugFix] Fix flash_attn import in `siglip2navit.py` (#29082)
Signed-off-by: Fanli Lin <fanli.lin@intel.com>
Fanli Lin committed
Nov 20, 2025 at 20:14 UTC
a2e9ebe9e242295a58e400835ef98a14b29c4fb0
1 file changed
+1
-1
vllm/model_executor/models/siglip2navit.py
+1
-1
@@ -191,7 +191,7 @@ def apply_rotary_pos_emb(
191
cos = cos.chunk(2, dim=-1)[0].contiguous()
192
sin = sin.chunk(2, dim=-1)[0].contiguous()
193
if is_flash_attn_backend and not current_platform.is_xpu():
194
- from flash_attn.layers.rotary import apply_rotary_emb
194
+ from vllm.vllm_flash_attn.layers.rotary import apply_rotary_emb
195
196
apply_rotary_emb_func = apply_rotary_emb
197
else: