Checkpoint for Luna's fine tuned model that I can refer back to.

Still halucinates and repeats sometimes, but way better than v1. Trained on FineTome, a massive reddit dataset, and scraped DM's from me and my friends. 
I think I'm on a great path, soon Luna will respond like a human.

Test out Luna's newest model - https://lunapotato.com/

Here's a snippet of a conversation I had with Luna in discord. 
![image/png](https://cdn-uploads.huggingface.co/production/uploads/6791917e9e215712a7bf9e46/nxax4rvwwhRNwN0_Px0iM.png)

With memory system turned off

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6791917e9e215712a7bf9e46/k8kl3jhw7xwIwteLXapHB.png)


Set parameters to this for best results:

![image/png](https://cdn-uploads.huggingface.co/production/uploads/6791917e9e215712a7bf9e46/3ramxCZKDxLpCsuz9n5zT.png)