metadata

base_model: unsloth/Llama-3.2-3B-Instruct-bnb-4bit
language:
  - en
license: apache-2.0
tags:
  - text-generation-inference
  - transformers
  - unsloth
  - llama
  - trl
  - sft
datasets:
  - yahma/alpaca-cleaned

🍷 Llama-3.2-3B-Instruct-Alpaca

This is a finetune of meta-llama/Llama-3.2-3B-Instruct.

It was trained on the yahma/alpaca-cleaned dataset using Unsloth.

This was my first fine tune and it's not done the best, but it is usable for small applications.

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "itsnebulalol/Llama-3.2-3B-Instruct-Alpaca"
messages = [{"role": "user", "content": "What is a large language model?"}]

tokenizer = AutoTokenizer.from_pretrained(model)
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
pipeline = transformers.pipeline(
    "text-generation",
    model=model,
    torch_dtype=torch.float16,
    device_map="auto",
)

outputs = pipeline(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])

This llama model was trained 2x faster with Unsloth and Huggingface's TRL library.