Instructions to use Billyshears/Janus_pro_finetune with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Billyshears/Janus_pro_finetune with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="Billyshears/Janus_pro_finetune")

# Load model directly
from transformers import MultiModalityCausalLM
model = MultiModalityCausalLM.from_pretrained("Billyshears/Janus_pro_finetune", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Billyshears/Janus_pro_finetune with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Billyshears/Janus_pro_finetune"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Billyshears/Janus_pro_finetune",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/Billyshears/Janus_pro_finetune

SGLang

How to use Billyshears/Janus_pro_finetune 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 "Billyshears/Janus_pro_finetune" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Billyshears/Janus_pro_finetune",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

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 "Billyshears/Janus_pro_finetune" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Billyshears/Janus_pro_finetune",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use Billyshears/Janus_pro_finetune with Docker Model Runner:
```
docker model run hf.co/Billyshears/Janus_pro_finetune
```

Janus-Pro-7B ScienceQA SFT checkpoint

This private checkpoint is the stage-1 ScienceQA supervised-fine-tuning result used by the Janus thesis reproduction in KFCCrazzzyThursday/Janus_pro_finetune.

Base model: deepseek-ai/Janus-Pro-7B
Training view: 5,678 ScienceQA image-only examples with official solutions
Training: full-parameter BF16, 267 optimizer steps, four GPUs
Intended next stage: TQA GRPO

The thesis did not report the stage-1 SFT hyperparameters, so this checkpoint uses the documented reproduction assumptions. See configs/paper.yaml and reproducibility/paper_audit.md in the code repository before interpreting or redistributing the model.

This repository contains model artifacts only. It intentionally contains no dataset images, API keys, judge responses, or training logs.

Downloads last month: 25

Safetensors

Model size

7B params

Tensor type

BF16

Inference Providers NEW

Image-Text-to-Text

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Billyshears/Janus_pro_finetune

Base model

deepseek-ai/Janus-Pro-7B

Finetuned

(37)

this model