Hacker News Reader

There's no GGUF available, but the process shouldn't be too hard from the provided .ckpt PyTorch checkpoint.

404

After downloading, it looks like it uses a custom "talkie" model architecture that is not supported at least by llama.cpp.

Edit: This one has now been deleted as well for some reason...

Update: here is a patched llama.cpp and quantized model for desktop use: https://github.com/solwyc/talkie-1930-13b-it-q5