Instructions to use Someshfengde/llama-3-instruction-tuned-AIMO with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Someshfengde/llama-3-instruction-tuned-AIMO with Transformers:

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

pipe = pipeline("text-generation", model="Someshfengde/llama-3-instruction-tuned-AIMO")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Someshfengde/llama-3-instruction-tuned-AIMO")
model = AutoModelForCausalLM.from_pretrained("Someshfengde/llama-3-instruction-tuned-AIMO")
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 Someshfengde/llama-3-instruction-tuned-AIMO with vLLM:

Install from pip and serve model

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

Use Docker

docker model run hf.co/Someshfengde/llama-3-instruction-tuned-AIMO

SGLang

How to use Someshfengde/llama-3-instruction-tuned-AIMO 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 "Someshfengde/llama-3-instruction-tuned-AIMO" \
    --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": "Someshfengde/llama-3-instruction-tuned-AIMO",
		"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 "Someshfengde/llama-3-instruction-tuned-AIMO" \
        --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": "Someshfengde/llama-3-instruction-tuned-AIMO",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use Someshfengde/llama-3-instruction-tuned-AIMO with Docker Model Runner:
```
docker model run hf.co/Someshfengde/llama-3-instruction-tuned-AIMO
```

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Instruction Tuning LLAMA3

This repo uses the torchtune for instruction tuning the llama3 pretrained model on mathematical tasks using LORA.

Wandb report link

https://wandb.ai/som/torchtune_llama3?nw=nwusersom

Instruction_tuned Model

https://huggingface.co/Someshfengde/llama-3-instruction-tuned-AIMO

Original metallama model

https://huggingface.co/meta-llama/Meta-Llama-3-8B

For running this project

> pip install poetry 
> poetry install

Further commands over shell terminal

To download the model

tune download meta-llama/Meta-Llama-3-8B \
--output-dir llama3-8b-hf \
--hf-token <HF_TOKEN>

To start instruction tuning with lora and torchtune

tune run lora_finetune_single_device --config ./lora_finetune_single_device.yaml

To quantize the model

tune run quantize --config ./quantization_config.yaml

To generate inference from model.

tune run generate --config ./generation_config.yaml \
prompt="what is 2 + 2."

Dataset used

https://huggingface.co/datasets/Someshfengde/AIMO_dataset

Evaluations

To run evaluations

tune run eleuther_eval --config ./eval_config.yaml

Someshfengde
/

llama-3-instruction-tuned-AIMO