GGML models for CPU inference
#1
by
Vetehinen
- opened
Thank you for the amazing work TurkuNLP!
For curious people with less compute power at their disposal, I uploaded bloomz.cpp GGML 4 bit quantized versions of 3B, 8B and 13B models.
No GPU required, only ability to compile and run bloomz.cpp and some RAM to load model.
https://huggingface.co/Vetehinen/gpt3-finnish-3B-GGML
https://huggingface.co/Vetehinen/gpt3-finnish-8B-GGML
https://huggingface.co/Vetehinen/gpt3-finnish-13B-GGML
hi