Project Synopsis:

• Experiment Overview:

  • Conducting a multifaceted test of LoRa's capabilities in AI image generation.

  • Three phases: Onsite Generator, Auto1111, and Comfyui exploration.

• Phase 1: Onsite Generator with Checkpoints

  • Trained LoRa model with 100 flower images across 10 diverse checkpoints.

  • Utilized onsite generator exclusively to observe LoRa's adaptation and image generation.

  • Aimed to establish baseline performance in a controlled environment.

• Phase 2: Auto1111 Exploration

  • Replicated Phase 1 parameters on Auto1111 platform for comparison.

  • Evaluated LoRa's performance and output quality in a different AI environment.

• Phase 3: Testing with Comfyui

  • Conducted Phase 1 parameters on Comfyui platform to further explore LoRa's capabilities.

  • Assessed how different AI platforms influence image generation outcomes.

LoRa Details:

• LoRa model trained with 1000 flower images.

• No captions provided during training.

• Epoch set to 14.

• Hypothesis: Training LoRa with diverse flower images will produce usable outputs.

• Sharing LoRa model and dataset for public access and collaboration.