rt-bench
contains benchmark code to compare the following ML frameworks:
- ggml
- ggml on gpu
- onnx
- candle
Clone this repository with git-lfs
installed:
git clone https://github.com/bloopai/rt-bench
This should download necessary model & tokenizer files. You can now run benchmarks with:
cargo bench
The generated violin plots are stored under target/criterion/embedding
.