INT4 LoRA fine-tuning vs QLoRA: A user inquired about the variations in between INT4 LoRA wonderful-tuning and QLoRA in terms of precision and speed. Yet another member explained that QLoRA with HQQ requires frozen quantized weights, doesn't use tinnygemm, and makes use of dequantizing together with torch.matmulUpdate vision design to gpt-4o by Mik