I wish they would release the quantized versions in a safetensor format. Many frameworks can't load PTE and GGUF.