INT4 LoRA wonderful-tuning vs QLoRA: A user inquired about the discrepancies between INT4 LoRA fine-tuning and QLoRA in terms of accuracy and speed. Yet another member explained that QLoRA with HQQ involves frozen quantized weights, will not use tinnygemm, and makes use of dequantizing together with torch.matmul "Automation isn't really chan