base_model: unsloth/Meta-Llama-3.1-8B-bnb-4bit language:
- en license: apache-2.0 tags:
- text-generation-inference
- transformers
- unsloth
- llama
- trl
- sft datasets:
- mlabonne/FineTome-100k
QuantFactory/FineLlama-3.1-8B-GGUF
This is quantized version of mlabonne/FineLlama-3.1-8B created using llama.cpp
Original Model Card
🍷 FineLlama-3.1-8B
This is a finetune of meta-llama/Meta-Llama-3.1-8B made for my article "Fine-tune Llama 3.1 Ultra-Efficiently with Unsloth".
It was trained on 100k super high-quality samples from the mlabonne/FineTome-100k dataset.
Try the demo: https://huggingface.co/spaces/mlabonne/FineLlama-3.1-8B
🔎 Applications
This model was made for educational purposes. I recommend using Meta's instruct model for real applications.
⚡ Quantization
🏆 Evaluation
TBD.
💻 Usage
!pip install -qU transformers accelerate
from transformers import AutoTokenizer
import transformers
import torch
model = "mlabonne/FineLlama-3.1-8B"
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.
