README.md

---
license: apache-2.0
base_model:
- some1nostr/Ostrich-Llama-3-70B
pipeline_tag: text-generation
---

![Ostrich-70B](https://primal.b-cdn.net/media-cache?s=o&a=1&u=https%3A%2F%2Fm.primal.net%2FHyFP.png)

# Model Card for Ostrich

**Finetuned from model:** [Llama3-70B](https://huggingface.co/meta-llama/Meta-Llama-3-70B-Instruct)

# Pre Trained With

- Health, nutrition, medicinal herbs
- Fasting, faith, healing
- Nostr notes: bitcoin, libertarian views and other topics discussed on [Nostr](https://coracle.social/)
- Some controversial and non mainstream topics


# Uses

Get a nuanced view on things. The answers from this model are quite different than majority of models on "topics that matter" like healthy living, nutrition.
Libertarian, based and veracious, seeking truth and not caring about the mainstream mediocrity. 

I am using the model here as a ground truth for Nostr related questions: https://wikifreedia.xyz/based-llm-leaderboard/npub1nlk894teh248w2heuu0x8z6jjg2hyxkwdc8cxgrjtm9lnamlskcsghjm9c


# Warning

Users (both direct and downstream) should be aware of the risks, biases and limitations of the model.
The trainer, developer or uploader of this model does not assume any liability. Use it at your own risk.
There is no guarantee that the model will be of any use. It may hallucinate often. 
The answers of this model will be often unreliable and you should do your own research and compare outputs.

# Training Details

## Training Data

Mostly what makes sense to me. Some books, some videos, some banned videos from youtube. 

## Training Procedure

LLaMa-Factory is used to train using fsdp_qlora technique. 

The number in the filenames (like v43000) means the version. I take the training steps and add those as the version. Each source, book, note adds to the version.


# Contributions

Let me know about your non mainstream models. Send me your ideas about how to curate models with better wisdom.

If you think you can curate great models with the best wisdom but don't know how to do the tech part, contact me. We could fine tune a model together.

# Sponsored by

https://pickabrain.ai

You can chat with newer versions of this model on https://pickabrain.ai .