floppyx3-MEGAmodel

despite the name, it's just a lil guy

NOTE

It is extremely important that you download the base model with this (in the model tree), so that it can generate. Additionally, to see updates, check the community posts on the base model.

What is this thing?

The MEGAmodel is a ~1M parameter model that had a LoRA trained on a Raspberry Pi 5 with 8GB of RAM. Obviously, this was not a good idea. "Mega" (as in megabyte) refers to 1.00e+6 (a million.) So he's tiny. This model is the base model version. It outputs nonsense. It's a tiny random lil guy.

How do I use it?

Download the base model (shown in the model tree), and import it into your preferred tool. MAKE SURE YOUR TOOL HAS LoRA SUPPORT!! Then, apply this LoRA. If you want a slightly smarter model, download the min branch instead of the main branch. Chats should be formatted as:

### User:
<user prompt>
### Assistant:
<assistant response>

Even with the LoRA, this thing is still super dumb. Are you planning to fix it?

Yes. My next plan is to do a floppyx4. It will probably have a much bigger dataset (as this one is currently plateauing), and the model will be pre-merged (to minimize confusion on how to use). In that repo, there will also be seperate base and lora branches for developer use. (Read the community post by me for why I changed this section!)

Who the hell are you?

I'm Nilky (not milky). I make stupid stuff that makes no sense.

Don't you have anything else to do?

Yes. Yes, I do.

What does "floppyx3" mean?

The base model (but not the rest lol) fits on 3 3.5" 1.44 MB floppy disks.

How do I cite?

Cite as:

[Floppy×3 MEGAmodel LoRA](https://huggingface.co/NILKNARFGonzo/floppyx3-MEGAmodel) by [Nilky](https://huggingface.co/NILKNARFGonzo) is licensed under [CC BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/).