The-Mirror-Prismals-Company/Mira-AI-2025-1-21-Blue

How to run this model

1. 1.Download Python and Text Editor/IDE/CMD Terminal
2. 2.Download inference script, save it as a .py file
3. 3.run the .py from a terminal, and open up the link it gives you in a web browser.

Introducing Mira, a small language model focused on ethically sourced training data. Mira's corpus is publically avalible to download and train your own model on and is dual licensed under MIT and GPL, you can pick whatever suits the need of your project. Mira's corpus consists of 28k words of my own inputs when talking to various large language models. Outputs from the language models were not used, only (my) user inputs, quotes from LLMs consist of a very small portion of sections where they are quoted in an input. I estimate less than 5% of the corpus is Ai generated, but none of the corpus contains outputs from another companies language model, so you can use this training data in your projects without the violating ethical or legal considerations. The Corpus is modular and very small, meaning you can train your own model and add to the corpus to upgrade your model. At only 124M parameters, almost anyone can train a mira model.