[XPU] Support LoRA via torch.compile on XPU platform (#36962)
Signed-off-by: chzhang <chaojun.zhang@intel.com>
This commit is contained in:
@@ -167,7 +167,7 @@ class XPUPlatform(Platform):
|
|||||||
cache_config.block_size = 64
|
cache_config.block_size = 64
|
||||||
|
|
||||||
# lazy import to avoid circular import
|
# lazy import to avoid circular import
|
||||||
from vllm.config import CompilationMode, CUDAGraphMode
|
from vllm.config import CUDAGraphMode
|
||||||
|
|
||||||
compilation_config = vllm_config.compilation_config
|
compilation_config = vllm_config.compilation_config
|
||||||
if compilation_config.compile_sizes is None:
|
if compilation_config.compile_sizes is None:
|
||||||
@@ -200,8 +200,6 @@ class XPUPlatform(Platform):
|
|||||||
"falling back to PIECEWISE graph mode on XPU platform."
|
"falling back to PIECEWISE graph mode on XPU platform."
|
||||||
)
|
)
|
||||||
|
|
||||||
if vllm_config.lora_config is not None:
|
|
||||||
compilation_config.mode = CompilationMode.NONE
|
|
||||||
# check and update parallel config
|
# check and update parallel config
|
||||||
parallel_config = vllm_config.parallel_config
|
parallel_config = vllm_config.parallel_config
|
||||||
# Only override worker_cls if it's still the default "auto"
|
# Only override worker_cls if it's still the default "auto"
|
||||||
|
|||||||
Reference in New Issue
Block a user