This is a port of BlinkDL/RWKV-LM to ggerganov/ggml. Besides the usual FP32, it supports FP16, quantized INT4, INT5 and INT8 inference. This project is focused on CPU, but cuBLAS is also supported.

The new CPU is now listed on AMD's website, as spotted by MEGAsizeGPU in a post on X (formerly Twitter), where you can find details of the specs. The Ryzen 5 5500X3D has the same amount of cache ...