Instructions to use ngxson/MiniThinky-v2-1B-Llama-3.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use ngxson/MiniThinky-v2-1B-Llama-3.2 with Transformers:

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

pipe = pipeline("text-generation", model="ngxson/MiniThinky-v2-1B-Llama-3.2")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("ngxson/MiniThinky-v2-1B-Llama-3.2")
model = AutoModelForCausalLM.from_pretrained("ngxson/MiniThinky-v2-1B-Llama-3.2")
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]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use ngxson/MiniThinky-v2-1B-Llama-3.2 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "ngxson/MiniThinky-v2-1B-Llama-3.2"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "ngxson/MiniThinky-v2-1B-Llama-3.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/ngxson/MiniThinky-v2-1B-Llama-3.2

SGLang

How to use ngxson/MiniThinky-v2-1B-Llama-3.2 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 "ngxson/MiniThinky-v2-1B-Llama-3.2" \
    --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": "ngxson/MiniThinky-v2-1B-Llama-3.2",
		"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 "ngxson/MiniThinky-v2-1B-Llama-3.2" \
        --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": "ngxson/MiniThinky-v2-1B-Llama-3.2",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use ngxson/MiniThinky-v2-1B-Llama-3.2 with Docker Model Runner:
```
docker model run hf.co/ngxson/MiniThinky-v2-1B-Llama-3.2
```

MiniThinky 1B

This is the newer checkpoint of MiniThinky-1B-Llama-3.2 (version 1), which the loss decreased from 0.7 to 0.5

Link to GGUF version: click here

Chat template is the same with llama 3, but the response will be as follow:

<|thinking|>{thinking_process}
<|answer|>
{real_answer}

IMPORTANT: System message

The model is very sensitive to system message. Make sure you're using this system message (system role) at the beginning of the conversation:

You are MiniThinky, a helpful AI assistant. You always think before giving the answer. Use <|thinking|> before thinking and <|answer|> before giving the answer.

Q&A

Hardware used to trained it?
I used a HF space with 4xL40S, trained for 5 hours (v1) and an additional of 6 hours (v2)

Benchmark?
I don't have time to do it alone. If you can help, please open a discussion!

Can it count number of "r" in "raspberry"?
Unfortunately no

Other things that I can tune?
Maybe lower temperature, or set top_k=1

TODO: include more info here + maybe do some benchmarks? (Plz add a discussion if you're interested)

Downloads last month: 150

Safetensors

Model size

1B params

Tensor type

F32

Model tree for ngxson/MiniThinky-v2-1B-Llama-3.2

Base model

meta-llama/Llama-3.2-1B-Instruct

Quantized

(371)

this model

Quantizations

7 models

Dataset used to train ngxson/MiniThinky-v2-1B-Llama-3.2

Spaces using ngxson/MiniThinky-v2-1B-Llama-3.2 16

Collection including ngxson/MiniThinky-v2-1B-Llama-3.2

MiniThinky: extra small reasoning models

Collection

My first trial to make reasoning models • 7 items • Updated Jan 15, 2025 • 2