The final versions of OpenLLaMa 3B and 7B have been released

[codesoap]

JFYI: Pretraining for the 3B and 7B models is complete:

• https://huggingface.co/openlm-research/open_llama_3b
• https://huggingface.co/openlm-research/open_llama_7b

PS: Training for a 13B model has also begun: https://huggingface.co/openlm-research/open_llama_13b_600bt

[codesoap]

The 13B model is now fully trained as well:

• https://huggingface.co/openlm-research/open_llama_13b

[limcheekin]

Hi there,

A viable alternative to this project:

• https://huggingface.co/VMware/open-llama-7b-open-instruct
• https://huggingface.co/VMware/open-llama-13b-open-instruct

Good day! :)

[codesoap]

@limcheekin I was not aware of those models. Thanks! I'm checking them out right now.

Though I must say, that I really liked the ### Input: option of OpenAlpaca. The open-instruct models don't seem to be trained for this.

[MarkusPfundstein]

hey guys! Great work! How much VRAM do we need for inference?