Instructions to use Amu/spin-phi2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Amu/spin-phi2 with Transformers:

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

pipe = pipeline("text-generation", model="Amu/spin-phi2", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Amu/spin-phi2", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Amu/spin-phi2", trust_remote_code=True)
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

vLLM

How to use Amu/spin-phi2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Amu/spin-phi2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Amu/spin-phi2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Amu/spin-phi2

SGLang

How to use Amu/spin-phi2 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 "Amu/spin-phi2" \
    --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": "Amu/spin-phi2",
		"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 "Amu/spin-phi2" \
        --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": "Amu/spin-phi2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Amu/spin-phi2 with Docker Model Runner:
```
docker model run hf.co/Amu/spin-phi2
```

outputs

This model is a fine-tuned version of microsoft/phi-2 using SPIN on ultrachat_200k dataset.

What's new

I think SPIN not only can use on a SFT model, but also it can use on a pretrained model. Therefore, I use SPIN on a pretrained model microsoft/phi-2. And I get a higher score better than origin pretrained model. You can check the open llm leaderboard.

But the ultrachat_200k dataset is a alignment dataset for sft model. I think there should use a alignment dataset for pretrained model.

I Think the best paradigm for training a conversational Large Language Model (LLM): pretrain -> dpo(spin) -> sft -> dpo(spin)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-07
train_batch_size: 1
eval_batch_size: 1
seed: 42
distributed_type: multi-GPU
num_devices: 8
gradient_accumulation_steps: 8
total_train_batch_size: 64
total_eval_batch_size: 8
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.1
num_epochs: 1

Framework versions

Transformers 4.37.0
Pytorch 2.1.2+cu121
Datasets 2.14.6
Tokenizers 0.15.2

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	61.68
AI2 Reasoning Challenge (25-Shot)	63.57
HellaSwag (10-Shot)	75.57
MMLU (5-Shot)	57.93
TruthfulQA (0-shot)	46.22
Winogrande (5-shot)	73.48
GSM8k (5-shot)	53.30

Downloads last month: 101

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for Amu/spin-phi2

Base model

microsoft/phi-2

Finetuned

(410)

this model

Collection including Amu/spin-phi2

RLXF

Collection

the best collection of RLXF model including RLHF, RLAIF etc. • 3 items • Updated Apr 7, 2025 • 1

Evaluation results

normalized accuracy on AI2 Reasoning Challenge (25-Shot)
test set Open LLM Leaderboard

63.570
normalized accuracy on HellaSwag (10-Shot)
validation set Open LLM Leaderboard

75.570
accuracy on MMLU (5-Shot)
test set Open LLM Leaderboard

57.930
mc2 on TruthfulQA (0-shot)
validation set Open LLM Leaderboard

46.220
accuracy on Winogrande (5-shot)
validation set Open LLM Leaderboard

73.480
accuracy on GSM8k (5-shot)
test set Open LLM Leaderboard

53.300