Hugging Face: πŸ€— Transformers adds native quantization: bitsandbytes and auto-gptq for PyTorch models | SignalBreak | SignalBreak