Bora 1

Bora 1 is an SLM created from scratch, designed to be a chatbot.

Using Kaggle's free GPU Quotas, this model was trained completely for free. Bora is a message to people that anyone at any status can achieve anything.

This model has 197,346,816 Parameters (197M).

Training Details

This model was trained in a Kaggle Jupyter Notebook. It utilized Kaggle's P100 GPU Accelerator. This model was trained with a single P100 GPU on about 1.5gb of data in about 90 hours.

All datasets used were from HuggingFace using the Datasets library.

Datasets used to pretrain:

  • nampdn-ai/mini-en
  • HuggingFaceTB/cosmopedia-100k

Datasets used to finetune:

  • HuggingFaceH4/SystemChat
  • HuggingFaceH4/no_robots
  • B-ai/DeepSeek-V3_Synthetic_Conversation_Dialogue

How to use

inference.py example usage

if __name__ == "__main__":
    checkpoint_path = "Bora1.pt"
    prompt = "Hello, how can I assist you today?"
    
    result = generate_text_from_checkpoint(
        checkpoint_path=checkpoint_path,
        prompt=prompt,
        max_length=100,
        temperature=0.7,
        top_k=50,
        top_p=0.9
    )
    
    # Print result as JSON
    print(json.dumps(result, indent=2))

Output example

{
  "status": "success",
  "generated_text": "Hello, how can I assist you today? I'm here to help with any questions or tasks you might have.",
  "prompt": "Hello, how can I assist you today?",
  "max_length": 100,
  "temperature": 0.7,
  "top_k": 50,
  "top_p": 0.9
}

Special Tokens

<user></user>
<assistant></assistant>
<system></system>
  • Messages separated by 1 newline.
  • Conversations separated by 2 newlines.

Founder's Bio

My name is Brandon Baek and I am a 15 years old Korean-American. Since I was 10, I was always fascinated with the idea of chatbots and AI. I remember trying to make chatbots in Scratch, then that slowly evolved to Python. For the last 5 years, I have been consistently working towards making a chatbot. When ChatGPT was first introduced, it only fueled my drive even further. My limit in resources due to my financial status has inspired me to create light-weight AI models that anyone could use, no matter the size of their computer or the density of their wallet. My dream is for AI's capabilities in supporting people to be democratized to everyone so that even the poorest people could have access to AI's advanced aid in many fields such as Medical, Legal, Finance, Education, Agriculture, and many more.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.
The model cannot be deployed to the HF Inference API: The model has no library tag.

Model tree for brandonbaek/Bora-1

Unable to build the model tree, the base model loops to the model itself. Learn more.

Datasets used to train brandonbaek/Bora-1

Collection including brandonbaek/Bora-1