Show HN: Fine-tune an 8B model on a 4 GB laptop GPU

Soup is a CLI tool that simplifies fine-tuning large language models on consumer GPUs by automating infrastructure setup through a single YAML configuration file. The latest version (v0.72.4) introduces layer streaming technology that enables preference-based fine-tuning methods like DPO and KTO on 4GB GPUs by keeping the frozen base model out of VRAM and feeding it to the GPU layer-by-layer. This approach eliminates the need for SSH access, cloud services, or complex configuration while maintaining performance comparable to standard training methods.