Skip to content

[https://nvbugs/5549081][fix] Fix device id assignment for some vision models #8070

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 2 commits into from
Oct 2, 2025
+6 −3
Merged

[https://nvbugs/5549081][fix] Fix device id assignment for some vision models #8070

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
ff72f6e
Use current device for some vision models
chang-l Sep 29, 2025
9debdad
Add comment
chang-l Oct 2, 2025
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
3 changes: 2 additions & 1 deletion tensorrt_llm/_torch/models/modeling_hyperclovax.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -726,7 +726,8 @@ def __init__(self, model_config: ModelConfig[PretrainedConfig]):
self.vision_config = self.pretrained_config.vision_config

model_path = self.pretrained_config._name_or_path
self.device = f"cuda:{model_config.mapping.rank}"
# TODO: use config.mapping.get_local_rank() instead
self.device = f"cuda:{torch.cuda.current_device()}"

hf_model_config = AutoConfig.from_pretrained(model_path,
trust_remote_code=True)
Expand Down
3 changes: 2 additions & 1 deletion tensorrt_llm/_torch/models/modeling_llama.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -998,7 +998,8 @@ def __init__(self, model_config: ModelConfig[Llama4Config], *args,
**kwargs):
super().__init__()
self.pretrained_config = model_config.pretrained_config
self.device = f"cuda:{model_config.mapping.rank}"
# TODO: use config.mapping.get_local_rank() instead
self.device = f"cuda:{torch.cuda.current_device()}"

self.dtype = self.pretrained_config.text_config.torch_dtype

Expand Down
3 changes: 2 additions & 1 deletion tensorrt_llm/_torch/models/modeling_llava_next.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -295,7 +295,8 @@ def __init__(self, model_config: ModelConfig[PretrainedConfig], *args,
super().__init__()
self.model_config = model_config
self.pretrained_config = model_config.pretrained_config
self.device = f"cuda:{model_config.mapping.rank}"
# TODO: use config.mapping.get_local_rank() instead
self.device = f"cuda:{torch.cuda.current_device()}"
model_path = self.pretrained_config._name_or_path

# Determine the actual local path for model files
Expand Down