Instructions to use lmsys/vicuna-33b-v1.3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use lmsys/vicuna-33b-v1.3 with Transformers:

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

pipe = pipeline("text-generation", model="lmsys/vicuna-33b-v1.3")

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("lmsys/vicuna-33b-v1.3")
model = AutoModelForCausalLM.from_pretrained("lmsys/vicuna-33b-v1.3")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use lmsys/vicuna-33b-v1.3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "lmsys/vicuna-33b-v1.3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "lmsys/vicuna-33b-v1.3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/lmsys/vicuna-33b-v1.3

SGLang

How to use lmsys/vicuna-33b-v1.3 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 "lmsys/vicuna-33b-v1.3" \
    --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": "lmsys/vicuna-33b-v1.3",
		"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 "lmsys/vicuna-33b-v1.3" \
        --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": "lmsys/vicuna-33b-v1.3",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use lmsys/vicuna-33b-v1.3 with Docker Model Runner:
```
docker model run hf.co/lmsys/vicuna-33b-v1.3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Vicuna Model Card

Model Details

Vicuna is a chat assistant trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.

Developed by: LMSYS
Model type: An auto-regressive language model based on the transformer architecture.
License: Non-commercial license
Finetuned from model: LLaMA.

Model Sources

Repository: https://github.com/lm-sys/FastChat
Blog: https://lmsys.org/blog/2023-03-30-vicuna/
Paper: https://arxiv.org/abs/2306.05685
Demo: https://chat.lmsys.org/

Uses

The primary use of Vicuna is research on large language models and chatbots. The primary intended users of the model are researchers and hobbyists in natural language processing, machine learning, and artificial intelligence.

How to Get Started with the Model

Command line interface: https://github.com/lm-sys/FastChat#vicuna-weights.
APIs (OpenAI API, Huggingface API): https://github.com/lm-sys/FastChat/tree/main#api.

Training Details

Vicuna v1.3 is fine-tuned from LLaMA with supervised instruction fine-tuning. The training data is around 125K conversations collected from ShareGPT.com. See more details in the "Training Details of Vicuna Models" section in the appendix of this paper.