Haus1

Something appears to have fallen through the cracks with ggml's new CUDA build flag. I wound up reverting it back to the last known good version to build with until...

Thanks, it's no longer an issue for me

If I'm not mistaken the reason for the prefix is because most models don't interpret the initial token correctly, so this was used to pad it. The value of the...

Is it loading the first 30 seconds and then refusing to buffer any further? If you hold down reload it'll let you clear the cache, that and clearing cookies may...

> > Is it loading the first 30 seconds and then refusing to buffer any further? If you hold down reload it'll let you clear the cache, that and clearing...

> > Is it loading the first 30 seconds and then refusing to buffer any further? If you hold down reload it'll let you clear the cache, that and clearing...

> Its the new Nvidia driver. Either rollback to the older version of it or sign out from the google and watch it without singing it. I dont know what...

> This is the log from the developer tools console [www.youtube.com-1725398704722.log](https://github.com/user-attachments/files/16856605/www.youtube.com-1725398704722.log) just a heads up, you may want to scrub any ips from this. groups do target people on here...

> > If that's still being blocked locally with extensions disabled it's entirely possible there's a bug or something else confounding the issue. > > I do use ublock origin,...

Does it still happen with --no-kv-offload? This may be part of a wider issue with AMD and their LLVM implementation for BF16 / FP16 being incorrect. edit: Ref https://github.com/llvm/llvm-project/pull/71470/commits/6e0ed50893c698cd94db1cad0a5274495007630c