Instructions to use adalberto-temp/Llama-3.2-3B-Instruct-GOLD with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use adalberto-temp/Llama-3.2-3B-Instruct-GOLD with Transformers:

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

pipe = pipeline("text-generation", model="adalberto-temp/Llama-3.2-3B-Instruct-GOLD")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("adalberto-temp/Llama-3.2-3B-Instruct-GOLD")
model = AutoModelForCausalLM.from_pretrained("adalberto-temp/Llama-3.2-3B-Instruct-GOLD")
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 adalberto-temp/Llama-3.2-3B-Instruct-GOLD with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "adalberto-temp/Llama-3.2-3B-Instruct-GOLD"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "adalberto-temp/Llama-3.2-3B-Instruct-GOLD",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/adalberto-temp/Llama-3.2-3B-Instruct-GOLD

SGLang

How to use adalberto-temp/Llama-3.2-3B-Instruct-GOLD 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 "adalberto-temp/Llama-3.2-3B-Instruct-GOLD" \
    --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": "adalberto-temp/Llama-3.2-3B-Instruct-GOLD",
		"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 "adalberto-temp/Llama-3.2-3B-Instruct-GOLD" \
        --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": "adalberto-temp/Llama-3.2-3B-Instruct-GOLD",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use adalberto-temp/Llama-3.2-3B-Instruct-GOLD with Docker Model Runner:
```
docker model run hf.co/adalberto-temp/Llama-3.2-3B-Instruct-GOLD
```

Model Card for Llama-3.2-3B-Instruct-GOLD

This model is a fine-tuned version of meta-llama/Llama-3.2-3B-Instruct on the adalberto-temp/gold-mix-dataset dataset. It has been trained using TRL.

Quick start

from transformers import pipeline

question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
generator = pipeline("text-generation", model="adalberto-temp/Llama-3.2-3B-Instruct-GOLD", device="cuda")
output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
print(output["generated_text"])

Training procedure

This model was trained with GOLD.

Framework versions

TRL: 0.25.1
Transformers: 4.57.1
Pytorch: 2.8.0
Datasets: 4.3.0
Tokenizers: 0.22.1

Citations

Cite GOLD as:

@misc{patino2025unlocking,
    title        = {{Unlocking On-Policy Distillation for Any Model Family}},
    author       = {Carlos Miguel Patiño and Kashif Rasul and Quentin Gallouédec and Ben Burtenshaw and Sergio Paniego and Vaibhav Srivastav and Thibaud Frere and Ed Beeching and Lewis Tunstall and Leandro von Werra and Thomas Wolf},
    year         = 2025,
    url          = {https://huggingface.co/spaces/HuggingFaceH4/general-on-policy-logit-distillation},
}

Cite TRL as:

@misc{vonwerra2022trl,
    title        = {{TRL: Transformer Reinforcement Learning}},
    author       = {Leandro von Werra and Younes Belkada and Lewis Tunstall and Edward Beeching and Tristan Thrush and Nathan Lambert and Shengyi Huang and Kashif Rasul and Quentin Gallou{\'e}dec},
    year         = 2020,
    journal      = {GitHub repository},
    publisher    = {GitHub},
    howpublished = {\url{https://github.com/huggingface/trl}}
}

Downloads last month: 3

Safetensors

Model size

3B params

Tensor type

F32

Model tree for adalberto-temp/Llama-3.2-3B-Instruct-GOLD

Base model

meta-llama/Llama-3.2-3B-Instruct

Finetuned

(1598)

this model

adalberto-temp
/

Llama-3.2-3B-Instruct-GOLD