Users are reporting difficulties in training LoRAs (Low-Rank Adaptation) with Moss, a tool that appears to struggle with learning specific characters or fine details. Despite using multiple datasets that have been successful with other tools like Krea 2, Anima, and Z-Image, the likeness in Moss does not seem to converge properly. This issue is reportedly being experienced by multiple users across various platforms, including Twitter, Reddit, and Discord, with some progress noted by the creator of Fizgig, though convergence remains insufficient. AI
IMPACT Potential issues with LoRA training in Moss could impact users' ability to fine-tune AI models for specific characters or details.
RANK_REASON User-reported issues with a specific AI tool's functionality.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
