NicoPlayZ9002/NCNano

Model Card for Model ID

NCNano sucks, unless you train it. Please train it.

Model Details

Model Description

NCNano is a model I made with the help of G***ni because I wanted to learn how LLMs work, and how to make one.

• Developed by: NicoPlayZ9002
• Funded by [optional]: Nobody! This is free.
• Shared by [optional]: I don't know what this means LOL
• Model type: Very small LLM and also it sucks
• Language(s) (NLP): [More Information Needed]
• License: [More Information Needed]
• Finetuned from model [optional]: [More Information Needed]

Model Sources [optional]

• Repository: You're at it!

Uses

It should be used for either training yourself or just messing around.

Direct Use

Please only use this for messing around. It sucks if you don't train it yourself. If you can train it, please do.

Downstream Use [optional]

LITERALLY nothing. Well, depends on if you train it or not. If you don't, this sucks but you can train it yourself! If you do, you might make the model a lot better.

Out-of-Scope Use

Code. This thing SUCKS at code. Honestly, it sucks at everything on my machine LOL

The limitations of having a mac...

Bias, Risks, and Limitations

I don't know. LOL -Nico

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

You should train the model yourself with train.py for better results! You can change which dataset you want my model to train on, how many steps it should train for, and more! Have fun, and go wild.

How to Get Started with the Model

Use the code below to get started with the model.

python generate.py (Prompt) (Maximum Tokens)

Just know that this is a "fill-in-the-blank" model (i guess), so it will not answer your questions.

Training Details

My training was using the first 5000 entrys in TinyStories, for 1000 steps.

Training Data

Loss was at about 1.37 by the end of training.

Training Procedure

train.py. Just use train.py.

Training Hyperparameters

• Training regime: fp32. Simple as that.

Testing Data

What?

[More Information Needed]

Summary

In summary, you can train the model yourself but if you don't, it sucks. It really sucks.

Model Examination [optional]

Genuinely what does this mean -Nico

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

How the hell am I meant to know this??? It's a local model...

Technical Specifications [optional]

Model Architecture and Objective

Idk bro.

Compute Infrastructure

What

Hardware

I trained it on a Mac Mini M4 with 16GB of RAM. Yikes.

Software

Python, and that's pretty much it. More specifically PyTorch.

Citation [optional]

I cite myself, I did it myself.

HOO boy that was really hard to write.