Skip to content

read max model len from hf config #94

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Aug 6, 2025
Merged

read max model len from hf config #94

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion multimodal/dashinfer_vlm/api_server/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -58,7 +58,7 @@ def add_context_args(parser):
choices=["cuda"],
help="device (Default: cuda)",
)
group.add_argument("--max_length", type=int, default=32000, help="model max length")
group.add_argument("--max_length", type=int, default=128000, help="model max length")
group.add_argument("--max_batch", type=int, default=128, help="max batch")
group.add_argument(
"--parallel_size",
Expand Down
6 changes: 5 additions & 1 deletion multimodal/dashinfer_vlm/api_server/server.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -117,12 +117,16 @@ def init():
context.set("tokenizer", tokenizer)
cuda_devices = [str(i) for i in range(tensor_parallel_size)]
compute_unit = "CUDA:" + ",".join(cuda_devices)
if hasattr(config, "max_position_embeddings"):
engine_max_len = min(config.max_position_embeddings, context.get("max_length"))
else:
engine_max_len = context.get("max_length")
# allspark model config
as_model_config = allspark.AsModelConfig(
model_name=model_name,
model_path=as_graph_path,
weights_path=as_weight_path,
engine_max_length=context.get("max_length"),
engine_max_length=engine_max_len,
engine_max_batch=context.get("max_batch"),
compute_unit=compute_unit,
enable_prefix_cache=True if context.get("enable_prefix_cache") else False,
Expand Down
4 changes: 2 additions & 2 deletions multimodal/dashinfer_vlm/vl_inference/runtime/qwen_vl.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -267,8 +267,8 @@ def __init__(
"QWEN2-AL": "audio_attention_mask",
"GUMMY-AL": "audio_attention_mask",
}
self.max_input_len = int(getenv("DS_LLM_MAX_IN_TOKENS", "20000"))
self.max_total_len = int(getenv("DS_LLM_MAX_TOKENS", "32000"))
self.max_total_len = as_config.engine_max_length
self.max_input_len = self.max_total_len - 1

"""
支持input_tokens格式:
Expand Down