Instructions to use savanladani/week2-llama3.2-1B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use savanladani/week2-llama3.2-1B with Transformers:

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

pipe = pipeline("text-generation", model="savanladani/week2-llama3.2-1B")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("savanladani/week2-llama3.2-1B")
model = AutoModelForCausalLM.from_pretrained("savanladani/week2-llama3.2-1B", 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=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Inference
Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use savanladani/week2-llama3.2-1B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "savanladani/week2-llama3.2-1B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "savanladani/week2-llama3.2-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/savanladani/week2-llama3.2-1B

SGLang

How to use savanladani/week2-llama3.2-1B 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 "savanladani/week2-llama3.2-1B" \
    --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": "savanladani/week2-llama3.2-1B",
		"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 "savanladani/week2-llama3.2-1B" \
        --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": "savanladani/week2-llama3.2-1B",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use savanladani/week2-llama3.2-1B with Docker Model Runner:
```
docker model run hf.co/savanladani/week2-llama3.2-1B
```

Model Overview

This model is a fine-tuned variant of Llama-3.2-1B, leveraging ORPO (Optimized Regularization for Prompt Optimization) for enhanced performance. It has been fine-tuned using the mlabonne/orpo-dpo-mix-40k dataset as part of the Finetuning Open Source LLMs Course - Week 2 Project.

Intended Use

This model is optimized for general-purpose language tasks, including text parsing, understanding contextual prompts, and enhanced interpretability in natural language processing applications.

Evaluation Results

The model was evaluated on the following benchmarks, with the following performance metrics:

Tasks	Version	Filter	Metric		Value		Stderr
eq_bench	2.1	none	eqbench	↑	1.5355	±	0.9174
		none	percent_parseable	↑	16.9591	±	2.8782
hellaswag	1	none	acc	↑	0.4812	±	0.0050
		none	acc_norm	↑	0.6467	±	0.0048
ifeval	4	none	inst_level_loose_acc	↑	0.3993	±	N/A
		none	inst_level_strict_acc	↑	0.2974	±	N/A
		none	prompt_level_loose_acc	↑	0.2754	±	0.0192
		none	prompt_level_strict_acc	↑	0.1848	±	0.0167
tinyMMLU	0	none	acc_norm	↑	0.3996	±	N/A

Key Features

Model Size: 1 Billion parameters
Fine-tuning Method: ORPO
Dataset: mlabonne/orpo-dpo-mix-40k

Downloads last month: 3

Safetensors

Model size

1B params

Tensor type

BF16

Model tree for savanladani/week2-llama3.2-1B

Base model

meta-llama/Llama-3.2-1B

Finetuned

(935)

this model

Dataset used to train savanladani/week2-llama3.2-1B

Evaluation results

EQ-Bench (0-Shot) on mlabonne/orpo-dpo-mix-40k
self-reported

1.536