xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-21 02:55:29 +08:00
Code Issues Packages Projects Releases Wiki Activity

FIxing the AMD test failures caused by PR#16457 (#17511)

Browse Source 
Signed-off-by: Alexei V. Ivanov <alexei.ivanov@amd.com>
This commit is contained in:
Alexei-V-Ivanov-AMD 2025-05-01 01:23:07 -05:00 committed by GitHub
parent afb4429b4f
commit 7ab643e425
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 1 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
docker/Dockerfile.rocm
View File

@ -118,4 +118,3 @@ ENV TOKENIZERS_PARALLELISM=false
ENV HIP_FORCE_DEV_KERNARG=1 ENV HIP_FORCE_DEV_KERNARG=1
CMD ["/bin/bash"] CMD ["/bin/bash"]

2
vllm/model_executor/layers/rotary_embedding.py
View File

@ -80,7 +80,7 @@ def _apply_rotary_emb(x: torch.Tensor, cos: torch.Tensor, sin: torch.Tensor,
is_neox_style: Whether to use the Neox-style or GPT-J-style rotary is_neox_style: Whether to use the Neox-style or GPT-J-style rotary
positional embeddings. positional embeddings.
""" """
if current_platform.is_cuda_alike(): if current_platform.is_cuda():
return apply_rotary_emb(x.unsqueeze(0), cos, sin, return apply_rotary_emb(x.unsqueeze(0), cos, sin,
not is_neox_style).squeeze(0) not is_neox_style).squeeze(0)
else: else: