{"version":"1.0","workflow_uuid":"0f8a00ca-54af-11f1-9bc6-00163e2b0d79","workflow_title":"QLoRA — Memory-Efficient Fine-Tuning for Quantized LLMs","install_contract":{"version":"1.0","installReady":false,"title":"QLoRA — Memory-Efficient Fine-Tuning for Quantized LLMs","summary":"QLoRA enables fine-tuning of large language models on consumer GPUs by backpropagating gradients through a frozen 4-bit quantized model into Low-Rank Adapters. It reduces memory requirements enough to fine-tune a 65B parameter model on a single 48GB GPU while preserving full 16-bit performance.","assetType":"Configs","pageUrl":"https://tokrepo.com/en/workflows/asset-0f8a00ca","sourceUrl":"https://github.com/artidoro/qlora.git","intendedFor":[],"firstActions":[],"agentFirstSteps":[],"targetPaths":[],"verification":[],"startingPoints":[],"example":"","successOutcome":"","boundaries":[],"askUserIf":["the current workspace stack cannot be matched to a safe upstream template","the target path is not the project root, or an existing file should be merged instead of overwritten"]}}