portal-vlm: qwen3vl-lora
Fresh-training reference for the same-family porting row: from-scratch grounding LoRA (r16/a32, LM sites) on Qwen3-VL-4B, native 0-1000 dialect.
Part of the portal-vlm release - an independent replication of Ramp Labs' PorTAL (portable task adapters via hypernet-generated LoRA) extended to vision-language models on GUI grounding.
| ScreenSpot-v2 overall | web split | |
|---|---|---|
| this artifact | 92.5% | 90.2% |
- Base model:
Qwen/Qwen3-VL-4B-Instruct@ebb281ec70b05090aa6165b016eac8ec08e71b17 - Training config:
configs/qwen3vl_lora.yaml - Eval record:
results/qwen3vl_lora.json
Reproduce this row without training
git clone https://github.com/robbym-dev/portal-vlm && cd portal-vlm && uv sync
uv run python scripts/eval.py --config configs/qwen3vl_lora.yaml --adapter hf:manihani4/portal-vlm-qwen3vl-lora
Standard PEFT LoRA adapter - also loadable directly with peft.PeftModel.from_pretrained on the pinned base model.
- Downloads last month
- 8
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Model tree for manihani4/portal-vlm-qwen3vl-lora
Base model
Qwen/Qwen3-VL-4B-Instruct