Instructions to use Skywork/Skywork-R1V2-38B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Skywork/Skywork-R1V2-38B with Transformers:

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

pipe = pipeline("image-text-to-text", model="Skywork/Skywork-R1V2-38B", trust_remote_code=True)
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)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Skywork/Skywork-R1V2-38B", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use Skywork/Skywork-R1V2-38B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Skywork/Skywork-R1V2-38B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Skywork/Skywork-R1V2-38B",
		"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/Skywork/Skywork-R1V2-38B

SGLang

How to use Skywork/Skywork-R1V2-38B 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 "Skywork/Skywork-R1V2-38B" \
    --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": "Skywork/Skywork-R1V2-38B",
		"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 "Skywork/Skywork-R1V2-38B" \
        --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": "Skywork/Skywork-R1V2-38B",
		"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 Skywork/Skywork-R1V2-38B with Docker Model Runner:
```
docker model run hf.co/Skywork/Skywork-R1V2-38B
```

no think tokens generated by model (only chat template)

by LeePapa - opened May 31, 2025

Discussion

LeePapa

May 31, 2025

•

edited May 31, 2025

How can I parse the reasoning when there is no token generated by the model?

The model's generated output (the content field value) starts directly with the reasoning ("Okay, so the problem is asking...") without outputting a "think" tag itself. The "think" tag is only part of the prompt constructed by the template (conversation.py / tokenizer_config.json), not part of the text the model generates.

Critically, the model does not generate a "/think" tag which means that there is no delimiter token that can be used to identify when the thinking finishes and transitions to the actual response.

The result is the thinking text and final response are merged together and not seperated. Do you have a suggestion on how to solve this? (or maybe you are working on R1V3 using Qwen3-32B? :-] )

OrlandoHugBot

Skywork org Jun 3, 2025

•

edited Jun 3, 2025

Hi, thanks for raising this! Actually, the model does generate the "/think" tag in its prediction to indicate the end of the reasoning process. The "think" tag is indeed part of the chat template defined in tokenizer_config.json or conversation.py, so it’s prepended during prompt construction. However, the "/think" is expected to be generated by the model itself, serving as a clear delimiter between the internal reasoning and the final answer.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment