GGMLSharp is a API for C# to use ggml.
ggml is a wonderful C-language machine-learning tool, and now you can use it with C#.
GGMLSharp contains all ggml shared libs and some demos.

• Written in C# only
• Only depends on ggml
• DotNet 462 Support!
• All Demos can use safe code only!

mnist_cpu is a basic demo for learning how to use GGMLSharp. It contains two Linears.

mnist_cnn is a demos show how to use convolution. In this demo, there are two conv2d and pool max.

mnist_train is a demo shows how to train a model. The model is same as mnist_cpu.

simple_backend shows how to use GGMLSharp with cuda. In this demo, you shold take ggml.dll for cuda. You can get it with the help of ggml or you can download it from llama.cpp.

magika is a useful tool from google. It can help to get the style of a file in high speed.

Converter is a useful tool for converting llm models from bin/ckpt/safetensors to gguf without any python environment.

ModelLoader is a tool for loading safetensors or pickle file directly from binary data. This demo can help to learn how to read a model file without any help of python.

SAM (Segment Anything Model) can help us seprate things from an image.

TestOpt is a basic demo for optimizar.

Yolov3Tiny is a Demo shows how to implement YOLO object detection with ggml using pretrained model. The weight have been converted to gguf.