text-generation-inference

feat(server): Reworking the quantization script so it's still universal (not llama specific)

#587

Merged