text-generation-inference

cece8635 - Fixing rocm gptq by using triton code too (renamed cuda into triton). (#2691)