Project Synopsis:<ul><li>Experiment Overview:<ul><li>Conducting a multifaceted test of LoRa's capabilities in AI image generation.</li><li>Three phases: Onsite Generator, Auto1111, and Comfyui exploration.</li></ul></li><li>Phase 1: Onsite Generator with Checkpoints<ul><li>Trained LoRa model with 100 flower images across 10 diverse checkpoints.</li><li>Utilized onsite generator exclusively to observe LoRa's adaptation and image generation.</li><li>Aimed to establish baseline performance in a controlled environment.</li></ul></li><li>Phase 2: Auto1111 Exploration<ul><li>Replicated Phase 1 parameters on Auto1111 platform for comparison.</li><li>Evaluated LoRa's performance and output quality in a different AI environment.</li></ul></li><li>Phase 3: Testing with Comfyui<ul><li>Conducted Phase 1 parameters on Comfyui platform to further explore LoRa's capabilities.</li><li>Assessed how different AI platforms influence image generation outcomes.</li></ul></li></ul>LoRa Details:<ul><li>LoRa model trained with 1000 flower images.</li><li>No captions provided during training.</li><li>Epoch set to 14.</li><li>Hypothesis: Training LoRa with diverse flower images will produce usable outputs.</li><li>Sharing LoRa model and dataset for public access and collaboration.</li></ul>

Files with SHA256: c9a10aed3edb5d9369e30283f2b321b2ed2e387382affe6930e2a2106d7e59bf