mirror of
https://git.datalinker.icu/vllm-project/vllm.git
synced 2025-12-14 18:15:01 +08:00
[Bugfix] Fix pre-commit (#29601)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
parent
51906c8c55
commit
e6d4f3c254
@ -237,9 +237,8 @@ def main():
|
|||||||
|
|
||||||
for i, prompt_data in enumerate(test_prompts):
|
for i, prompt_data in enumerate(test_prompts):
|
||||||
print(
|
print(
|
||||||
f"\nRunning prompt {i + 1}/{len(test_prompts)}: {
|
f"\nRunning prompt {i + 1}/{len(test_prompts)}: "
|
||||||
prompt_data['description']
|
f"{prompt_data['description']}"
|
||||||
}"
|
|
||||||
)
|
)
|
||||||
|
|
||||||
output_str = run_chat_completion(
|
output_str = run_chat_completion(
|
||||||
|
|||||||
@ -35,13 +35,16 @@ class EagleCudaGraphManager:
|
|||||||
self.compilation_config = vllm_config.compilation_config
|
self.compilation_config = vllm_config.compilation_config
|
||||||
assert self.compilation_config is not None
|
assert self.compilation_config is not None
|
||||||
|
|
||||||
|
cudagraph_mode: CUDAGraphMode
|
||||||
if self.compilation_config.cudagraph_mode is None:
|
if self.compilation_config.cudagraph_mode is None:
|
||||||
self.cudagraph_mode = CUDAGraphMode.NONE
|
cudagraph_mode = CUDAGraphMode.NONE
|
||||||
else:
|
else:
|
||||||
self.cudagraph_mode = self.compilation_config.cudagraph_mode
|
cudagraph_mode = self.compilation_config.cudagraph_mode
|
||||||
if self.cudagraph_mode == CUDAGraphMode.FULL:
|
if cudagraph_mode == CUDAGraphMode.FULL:
|
||||||
# NOTE(woosuk): For Eagle, we only use CUDA graphs for decode.
|
# NOTE(woosuk): For Eagle, we only use CUDA graphs for decode.
|
||||||
self.cudagraph_mode = CUDAGraphMode.FULL_DECODE_ONLY
|
cudagraph_mode = CUDAGraphMode.FULL_DECODE_ONLY
|
||||||
|
|
||||||
|
self.cudagraph_mode = cudagraph_mode
|
||||||
|
|
||||||
self.cudagraph_sizes = get_cudagraph_sizes(
|
self.cudagraph_sizes = get_cudagraph_sizes(
|
||||||
self.compilation_config.cudagraph_capture_sizes,
|
self.compilation_config.cudagraph_capture_sizes,
|
||||||
|
|||||||
Loading…
x
Reference in New Issue
Block a user