Zagora, Distributed fine-tuning platform on mixed GPUs over internet
Pipeline parallelism for mixed GPUs over internet, but unproven vs established frameworks.
Ask a question, and the planchette moves across the board spelling out the spirit's answer. Closely related to the classic Ouija board.
Fine-tuned 3B model as a Ouija board with crisis detection—clever execution, zero practical use.
AI hobbyists, entertainment/novelty seekers, developers experimenting with fine-tuned models
Replika (local variant) · Character.AI (open-source clones)
Pipeline parallelism for mixed GPUs over internet, but unproven vs established frameworks.
Fine-tuned 0.6B model keeps agent knowledge graphs clean using dynamic text rules.
Galaxy classification model, but model card has mostly empty fields.
Eval-synthesize-train loop automates custom model development better than manual fine-tuning.
Fine-tuned 3B Qwen matches Haiku on jokes, validating small models for constrained agent tasks.
Unified memory trick lets a 2B model beat 12B; trains on MacBook with zero cloud costs.