Instructions to use UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3 with Transformers:

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

pipe = pipeline("text-generation", model="UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3")
model = AutoModelForCausalLM.from_pretrained("UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3", device_map="auto")
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3

SGLang

How to use UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3 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 "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3" \
    --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": "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3" \
        --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": "UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3 with Docker Model Runner:
```
docker model run hf.co/UCLA-AGI/zephyr-7b-sft-full-SPIN-iter3
```
Browse Quantizations to use this model in llama.cpp, Ollama, LM Studio, or any compatible app.

Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models (https://arxiv.org/abs/2401.01335)

zephyr-7b-sft-full-spin-iter3

This model is a self-play fine-tuned model at iteration 3 from alignment-handbook/zephyr-7b-sft-full using synthetic data based on on the HuggingFaceH4/ultrachat_200k dataset.

Model Details

Model Description

Model type: A 7B parameter GPT-like model fine-tuned on synthetic datasets.
Language(s) (NLP): Primarily English
License: MIT
Finetuned from model: alignment-handbook/zephyr-7b-sft-full (based on mistralai/Mistral-7B-v0.1)

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 1e-07
train_batch_size: 8
seed: 42
distributed_type: multi-GPU
num_devices: 8
total_train_batch_size: 64
optimizer: RMSProp
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 2.0

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	63.70
ARC (25-shot)	66.13
HellaSwag (10-shot)	85.85
MMLU (5-shot)	61.51
TruthfulQA (0-shot)	57.89
Winogrande (5-shot)	76.64
GSM8K (5-shot)	34.19

Citation

@misc{chen2024selfplay,
      title={Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models}, 
      author={Zixiang Chen and Yihe Deng and Huizhuo Yuan and Kaixuan Ji and Quanquan Gu},
      year={2024},
      eprint={2401.01335},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}