Fine-tuning a 1.7B model at 3.2 GB VRAM — building FineTune Studio

How FineTune Studio runs QLoRA fine-tuning end to end in a browser, on hardware a student can actually afford, with honest base-vs-tuned evaluation.