Instructions to use shaw2037/Llama-3.2-3B-Instruct-Reasoning with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use shaw2037/Llama-3.2-3B-Instruct-Reasoning with Transformers:

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

pipe = pipeline("text-generation", model="shaw2037/Llama-3.2-3B-Instruct-Reasoning")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("shaw2037/Llama-3.2-3B-Instruct-Reasoning")
model = AutoModelForCausalLM.from_pretrained("shaw2037/Llama-3.2-3B-Instruct-Reasoning", 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=256)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use shaw2037/Llama-3.2-3B-Instruct-Reasoning with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "shaw2037/Llama-3.2-3B-Instruct-Reasoning"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "shaw2037/Llama-3.2-3B-Instruct-Reasoning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/shaw2037/Llama-3.2-3B-Instruct-Reasoning

SGLang

How to use shaw2037/Llama-3.2-3B-Instruct-Reasoning 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 "shaw2037/Llama-3.2-3B-Instruct-Reasoning" \
    --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": "shaw2037/Llama-3.2-3B-Instruct-Reasoning",
		"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 "shaw2037/Llama-3.2-3B-Instruct-Reasoning" \
        --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": "shaw2037/Llama-3.2-3B-Instruct-Reasoning",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use shaw2037/Llama-3.2-3B-Instruct-Reasoning with Docker Model Runner:
```
docker model run hf.co/shaw2037/Llama-3.2-3B-Instruct-Reasoning
```

LLaMA 3B Instruct Reasoning Model

This model is a fine-tuned version of LLaMA 3B Instruct, optimized for reasoning tasks such as step-by-step problem solving and logical question answering.

The model was fine-tuned using LoRA (PEFT) and later merged into the base model to create a fully standalone model.

Base Model

meta-llama/Llama-3-3b-instruct

Model Details

Architecture: LLaMA 3B
Fine-tuning method: LoRA (merged)
Task: Causal Language Modeling
Use case: Reasoning / instruction-following

Features

Improved step-by-step reasoning
Better structured answers
Enhanced instruction following
Suitable for logical tasks

Training Details

This model was fine-tuned on a reasoning dataset from Hugging Face using LoRA.

The LoRA weights were merged with the base model to produce a standalone model for easier deployment and usage.

How to Use

from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "shaw2037/Llama-3.2-3B-Instruct-Reasoning"

tokenizer = AutoTokenizer.from_pretrained(model_id)

model = AutoModelForCausalLM.from_pretrained(
    model_id,
    device_map="auto"
)

prompt = "Solve step by step: If 2x + 3 = 11, what is x?"

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=200,
    
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Limitations

May produce incorrect reasoning steps.

Can hallucinate in complex scenarios.

Not guaranteed to be mathematically perfect.

Intended Use

Suitable for:

reasoning experiments

educational projects

LLM research

Not suitable for:

medical advice

legal advice

financial decisions

safety-critical applications

Install dependencies

pip install transformers accelerate torch

Downloads last month: 11

Safetensors

Model size

3B params

Tensor type

BF16

Model tree for shaw2037/Llama-3.2-3B-Instruct-Reasoning

Base model

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

Finetuned

(2030)

this model