kaggle notebook with LlamaCpp + OpenOrca-Zephyr-7B model

[wiiiktor]

https://www.kaggle.com/wiiiktor/llamacpp-demo/edit

[wiiiktor]

Here you can find a Colab notebook, where I loaded a 4bit model, so it works even on a V100 gpu, which is always available (not like A100, which you can access from time to time), BUT I have a "UnsupportedOperation: fileno" message there; feedback on this would be welcome, so that I can fix and republish this notebook:

https://colab.research.google.com/drive/1D8VYz_Q1X7BSXj7vP326t_HKEbHnQ48_#scrollTo=ml8txfUavjY0

[Harsha-Nori]

@wiiiktor, probably a bit late to the update here, but this is due to a bug on the llamacpp-python side that we patched on our side (see this issue for details: #545). Upgrading guidance should fix this issue :). Notebook looks great -- cool to see everything running on Colab!