Mark Schmidt

### Potential Solution 3 Simply run a cron job to scrape for updates from the [XML pages](https://rfc2822.gitlab.io/fdroid-firefox/fdroid/repo/index.xml) for a list of repos and analyze everything in a queue. The amount...

Duplicate of #1383

Put it in a folder next to the rest of your models and then select it from the dropdown.  Note that the "Tuned" versions need to be prompted like:...

> I think it needs to be quantized to run at full context in 24gb. There are some 16bit versions here: [huggingface.co/vvsotnikov/stablelm-tuned-alpha-7b-16bit](https://huggingface.co/vvsotnikov/stablelm-tuned-alpha-7b-16bit) [huggingface.co/vvsotnikov/stablelm-tuned-alpha-3b-16bit](https://huggingface.co/vvsotnikov/stablelm-tuned-alpha-3b-16bit) These might work with load-in-8bit aswell? (Haven't...

Yes the quality / benchmarks are currently not good, to say the least. I would hold out for the next training checkpoint tbh.

It's definitely not your PR's fault. See https://github.com/Stability-AI/StableLM/issues/30

I added a table of contents, cleaned up some confusing language, and made the "tags" full words for better readability ("bare" not "B" for example).

It's your thing man, go ahead. :)

> Good Job! I also want to try running on Colab Pro You can make the changes above to the Colab linked in the readme. https://colab.research.google.com/github/oobabooga/AI-Notebooks/blob/main/Colab-TextGen-GPU.ipynb Just click "Show Code"...

Nothing needs to be changed. Use load_in_8bit depending on how much VRAM your instance has and what size model you use. Refer to the table below. Model | 16bit VRAM...