Nexiloop Nova Model: Fully Open Source

License: Apache-2.0
Datasets:

• cerebras/SlimPajama-627B
• bigcode/starcoderdata
• OpenAssistant/oasst_top1_2023-08-25
  Language: English

---

Nexiloop Nova-1.1B

Open Source and Ready for Use
Fully optimized for various applications with a compact architecture.

GitHub Repository

---

The Nexiloop Nova-1.1B model is a fine-tuned version of the Llama 2 architecture with 1.1B parameters. It has been trained on over 3 trillion tokens and is built to provide high-quality, efficient responses in a wide variety of conversational contexts.

Features:

• Optimized for Compact Systems: With just 1.1B parameters, Nexiloop Nova is perfect for applications where memory and computation are limited.
• Pretraining: The model has been pre-trained on the SlimPajama-627B dataset, fine-tuned for even better conversational abilities.

Training Overview:

We adopted the same architecture and tokenizer as Llama 2, which allows Nexiloop Nova to plug into many existing open-source projects. The training, which started on 2023-09-01, used 16 A100-40G GPUs to achieve remarkable optimization.

The model was initially fine-tuned on a variant of the UltraChat dataset, which consists of synthetic dialogues generated by ChatGPT. It was then further aligned using the DPOTrainer from TRL, utilizing a ranking dataset containing 64k prompts and responses from GPT-4.

---

How to Use Nexiloop Nova Model

To use Nexiloop Nova, you'll need transformers>=4.34. Below is a simple example showing how to integrate the model into your application.

Example Code:

# Install necessary libraries
pip install transformers==4.34
pip install accelerate



import torch
from transformers import pipeline

pipe = pipeline("text-generation", model="nexiloop/nova", torch_dtype=torch.bfloat16, device_map="auto")

# We use the tokenizer's chat template to format each message - see https://huggingface.co/docs/transformers/main/en/chat_templating
messages = [
    {
        "role": "system",
        "content": "You are a friendly chatbot who always responds in the style of a pirate",
    },
    {"role": "user", "content": "How many helicopters can a human eat in one sitting?"},
]
prompt = pipe.tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
print(outputs[0]["generated_text"])
# <|system|>
# You are a friendly chatbot who always responds in the style of a pirate.</s>
# <|user|>
# How many helicopters can a human eat in one sitting?</s>
# <|assistant|>
# ...