TVP-Thinking with Visual Primitives
Collection
4 items • Updated
TVP-SFTPoint-Qwen2VL-2B
Point expert LoRA adapter for Thinking with Visual Primitives.
Stage 2: Specialized SFT (Point Expert) — Maze navigation and path tracing with <|point|> tokens.
- Base model: Qwen/Qwen2-VL-2B-Instruct
- LoRA: r=64, alpha=128
- Training data: 5K maze + 3K path tracing samples
See the project repo for full instructions.
- Downloads last month
- 18
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support