feat: update model_specific_adjustment (#5344)

Co-authored-by: hebiao064 <hebiaobuaa@gmail.com>
This commit is contained in:
Yineng Zhang
2025-04-15 14:45:15 -07:00
committed by GitHub
parent e8f62b20ca
commit fa909dc3c4
4 changed files with 51 additions and 17 deletions

View File

@@ -569,7 +569,7 @@ def encode_video(video_path, frame_count_limit=None):
def load_image(
image_file: Union[Image.Image, str, bytes]
image_file: Union[Image.Image, str, bytes],
) -> tuple[Image.Image, tuple[int, int]]:
image = image_size = None
if isinstance(image_file, Image.Image):
@@ -1905,3 +1905,28 @@ def get_local_ip_by_remote() -> str:
return s.getsockname()[0]
except Exception:
raise ValueError(f"Can not get local ip")
def is_page_size_one(server_args):
return server_args.page_size == 1
def is_no_spec_infer_or_topk_one(server_args):
return server_args.speculative_eagle_topk is None or (
server_args.speculative_eagle_topk is not None
and server_args.speculative_eagle_topk == 1
and is_page_size_one(server_args)
)
def is_fa3_default_architecture(hf_config):
architectures = getattr(hf_config, "architectures", None)
if not isinstance(architectures, list) or not architectures:
return False
default_archs = {
"Qwen2ForCausalLM",
"Llama4ForConditionalGeneration",
"LlamaForCausalLM",
"MistralForCausalLM",
}
return architectures[0] in default_archs