Lyrcaxis

BPE Tokenizer: Multiple newlines doesn't merge into a single token

2

So, I found out that `\n\n` if appended by a character tokenizes as `['\n',\n']` (`[198, 198]`) instead of `['\n\n']` (`[271]`). (I'm using Llama3 for this example, but this extends to...

I made a PR with usage examples for Whisper.net as a server and in combination with LlamaSharp: https://github.com/SciSharp/LLamaSharp/pull/656 I thought people may benefit as I think it's better than the...

Whisper support

7

Hi, is anyone interested in Speech-To-Text support? (whisper.cpp) If someone is willing to create the native handles for it I'd be glad to contribute the mid-level & high-level bindings for...