[torch.compile] limit inductor threads and lazy import quant (#10482)

Signed-off-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
youkaichao
2024-11-20 18:36:33 -08:00
committed by GitHub
parent 2f77b6cfec
commit 388ee3de66
11 changed files with 178 additions and 64 deletions
-3
View File
@@ -19,9 +19,6 @@ if not current_platform.is_tpu() and not current_platform.is_hpu():
except ImportError as e:
logger.warning("Failed to import from vllm._C with %r", e)
if current_platform.is_rocm():
import vllm._rocm_C # noqa: F401
supports_moe_ops = False
with contextlib.suppress(ImportError):
import vllm._moe_C # noqa: F401