README.md

metadata

license: llama3.2
language:
  - en
inference: false
fine-tuning: false
tags:
  - nvidia
  - llama3.2
datasets:
  - nvidia/HelpSteer2
base_model: unsloth/Llama-3.2-3B-Instruct-bnb-4bit
pipeline_tag: text-generation
library_name: transformers

🍷 Llama-3.2-Nemotron-3B-Instruct

This is a finetune of meta-llama/Llama-3.2-3B-Instruct (specifically, unsloth/Llama-3.2-3B-Instruct-bnb-4bit).

It was trained on the nvidia/HelpSteer2 dataset, similar to nvidia/Llama-3.1-Nemotron-70B-Instruct-HF, using Unsloth.

💻 Usage

!pip install -qU transformers accelerate

from transformers import AutoTokenizer
import transformers
import torch

model = "itsnebulalol/Llama-3.2-Nemotron-3B-Instruct"
messages = [{"role": "user", "content": "How many r in strawberry?"}]

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.