Image-Text-to-Text
Transformers
Safetensors
English
qwen3_vl
navigation
vln
vln-ce
instruction-generation
visual-trajectory-prompting
qwen3-vl
acmmm-2026
conversational
Instructions to use TidalYang/VTInstructor-8b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use TidalYang/VTInstructor-8b with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="TidalYang/VTInstructor-8b") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("TidalYang/VTInstructor-8b") model = AutoModelForMultimodalLM.from_pretrained("TidalYang/VTInstructor-8b", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use TidalYang/VTInstructor-8b with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "TidalYang/VTInstructor-8b" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "TidalYang/VTInstructor-8b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/TidalYang/VTInstructor-8b
- SGLang
How to use TidalYang/VTInstructor-8b with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "TidalYang/VTInstructor-8b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "TidalYang/VTInstructor-8b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "TidalYang/VTInstructor-8b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "TidalYang/VTInstructor-8b", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use TidalYang/VTInstructor-8b with Docker Model Runner:
docker model run hf.co/TidalYang/VTInstructor-8b
Update README.md
Browse files
README.md
CHANGED
|
@@ -207,13 +207,6 @@ Scoring additionally needs `pycocoevalcap` and a Java runtime on `PATH` (for MET
|
|
| 207 |
|
| 208 |
See the GitHub README for rendering, VTP mask construction, SFT, GRPO, and evaluation.
|
| 209 |
|
| 210 |
-
## Limitations
|
| 211 |
-
|
| 212 |
-
- **Rendered trajectories required.** Inference consumes egocentric keyframes produced by the Stage-1 Habitat renderer. Reproducing the benchmark numbers therefore requires Matterport3D scenes (subject to the Matterport3D Terms of Use) and `habitat-sim`.
|
| 213 |
-
- **Long trajectories are truncated.** Frames are capped at `MAX_FRAMES` (default 32), taken from the start of the trajectory, so the tail of very long RxR episodes may be dropped.
|
| 214 |
-
- **DPC not released.** The DPC component of EDTC described in the paper is not part of the current code release; the rest of the pipeline is.
|
| 215 |
-
- **Version-sensitive.** The VP-Adapter injects into Qwen3-VL internals and is validated only against `transformers==4.57.6`.
|
| 216 |
-
- **English only**, indoor Matterport3D-style scenes only. Instructions are not verified for physical safety and should not drive a real robot without a human in the loop.
|
| 217 |
|
| 218 |
## Citation
|
| 219 |
|
|
|
|
| 207 |
|
| 208 |
See the GitHub README for rendering, VTP mask construction, SFT, GRPO, and evaluation.
|
| 209 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 210 |
|
| 211 |
## Citation
|
| 212 |
|