[BUG]: Bad Typo in LLama/Extensions/IContextParamsExtensions.cs ?

[SerialKicked]

Description

While i was looking around the code trying to figure a few issues compared to base LLama.cpp (memory consumption and repetition penalty doing weird things), I found this:

LLamaSharp/LLama/Extensions/IContextParamsExtensions.cs

Lines 50 to 51 in c1302ef

	result.type_k = @params.TypeK ?? GGMLType.GGML_TYPE_F16;
	result.type_k = @params.TypeV ?? GGMLType.GGML_TYPE_F16;

param.TypeV and param.TypeK are both assigned to result.type_k

I'm not exactly an expert in LLama.cpp / GGUF architecture, but it seems there's a mix-up.

Reproduction Steps

N/A

Environment & Configuration

• Operating system: Win11
• .NET runtime version: 8.0
• LLamaSharp version: 0.13
• CUDA version (if you are using cuda backend): 12
• CPU & GPU device: RTX3060 12GB

Known Workarounds

No response

[martindevans]

That's definitely a bug! If you'd like to make a PR to fix it that would be appreciated :)

[SerialKicked]

Sure thing, if you can wait until this week-end (otherwise, feel free to change the line, I won't mind.)