Fizgig 8.9.0 adds multi-GPU LoRA training across PCs and rented pods
shootthesound · reddit · 2026-10-23
Fizgig 8.9.0, the free open-source LoRA training tool, lets spare GPUs anywhere contribute to your training run:
- Experimental 2-GPU training: each GPU trains its own LoRA copy on its share of images, merging every 3 epochs; likeness matched a one-GPU run in testing, and a dual RTX 4090 pod cut 40% off each epoch
- Previews never pause training: a helper machine renders them while your GPU keeps going
- Move runs freely: migrate training to a laptop, another PC, or a RunPod/Vast.ai pod and bring it back anytime
- Home GPUs help pods: checkpoints stream back as they're saved; the pod can auto-delete once done
- Renting tab: lists free RunPod cards cheapest-per-run first
- Also: Krea 2 now trains with 2 GB system RAM instead of 25, and the Look Filter finds furry/anime/drawn faces
Free and open source on GitHub; update and restart all helper machines and pods.
More from Infra
- HarnessRouter open-sources a unified API to run Codex, Claude Code and more as agent backends — tom_doerr · 2026-10-23
- Local AI security checklist: pin Ollama ≥0.35.0, vLLM ≥0.31.0, purge LiteLLM 1.107.0 malware — AIFlow_ML · 2026-10-23
- Open-source Bonsai 2 9B gets 471% faster on Macs after one-week community optimization challenge — gajesh · 2026-10-23
- July 28 MCP release makes remote servers stateless, biggest revision since launch — bibryam · 2026-10-23
- MCP keynote: AAIF hits 305 members, Skills over MCP and Tasks/Events roadmap revealed — bibryam · 2026-10-23
- Only 8% of cloud vendors support OIDC — dev calls to kill long-lived API keys — cramforce · 2026-10-23