biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Hardware][CPU] using current_platform.is_cpu (#9536)

Browse Source
This commit is contained in:
wangshuai09
2024-10-22 15:50:43 +08:00
committed by GitHub
parent 0d02747f2e
commit 3ddbe25502
17 changed files with 60 additions and 64 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
vllm/attention/selector.py
View File

@@ -10,7 +10,7 @@ import vllm.envs as envs
from vllm.attention.backends.abstract import AttentionBackend
from vllm.logger import init_logger
from vllm.platforms import current_platform
from vllm.utils import STR_BACKEND_ENV_VAR, is_cpu, is_hip, is_openvino, is_xpu
from vllm.utils import STR_BACKEND_ENV_VAR, is_hip, is_openvino, is_xpu
logger = init_logger(__name__)
@@ -121,7 +121,7 @@ def get_attn_backend(
ROCmFlashAttentionBackend)
return ROCmFlashAttentionBackend
elif backend == _Backend.TORCH_SDPA:
assert is_cpu(), RuntimeError(
assert current_platform.is_cpu(), RuntimeError(
"Torch SDPA backend is only used for the CPU device.")
logger.info("Using Torch SDPA backend.")
from vllm.attention.backends.torch_sdpa import TorchSDPABackend
@@ -183,7 +183,7 @@ def which_attn_to_use(
if backend_by_env_var is not None:
selected_backend = backend_name_to_enum(backend_by_env_var)
if is_cpu():
if current_platform.is_cpu():
if selected_backend != _Backend.TORCH_SDPA:
logger.info("Cannot use %s backend on CPU.", selected_backend)
return _Backend.TORCH_SDPA