LoRA Genomic Evolution: GPU-aware genome paging for local expert models#280

Merged

joelteply merged 6 commits intomainfrom

feature/lora-genome-academy

Mar 2, 2026

Commits on Mar 2, 2026

GPU Memory Manager: unified VRAM coordination for all GPU consumers
joelteply
committed
gpu/stats command: TypeScript command scaffold + IPC mixin for GPU memory manager
joelteply
committed
Wire GPU memory tracking into all consumers: inference, TTS, renderer, genome paging
joelteply
committed
PEFT training memory safety: gradient checkpointing, OOM handling, GPU pressure guard
joelteply
committed
Sentinel real-time observability, LLM retry, academy pipeline resilience
joelteply
committed
Strip markdown fences in sentinel interpolation, doc refresh
joelteply
committed