Skip to content

docs: add DoRA to roadmap #1124

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

Open
GeorgePearse wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs: add DoRA to roadmap #1124

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
16 changes: 16 additions & 0 deletions docs/roadmap.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -280,6 +280,22 @@ GPU-accelerated preprocessing for NVIDIA hardware:
- Pipeline-parallel execution
- Best for high GPU:CPU ratio systems

### DoRA (Weight-Decomposed Low-Rank Adaptation)

**Status**: Under evaluation

[DoRA](https://github.com/NVlabs/DoRA) is a parameter-efficient fine-tuning method that improves upon LoRA by decomposing pre-trained weights into magnitude and direction components:
- **Superior to LoRA** especially at lower ranks, allowing reduced memory consumption
- **No inference overhead** - zero additional computational cost during inference
- **Enhanced training stability** through weight decomposition approach
- **ICML 2024 Oral** (top 1.5% acceptance rate)
- **HuggingFace PEFT integration** - supports Linear, Conv1d, Conv2d layers and quantized models

Potential applications for visdet:
- Efficient fine-tuning of vision transformer backbones (Swin, ViT)
- Adaptation of detection heads for domain-specific tasks
- Low-resource model customization with minimal trainable parameters

---

## Timeline Summary
Expand Down