cyberneticlibrary

Extract LoRA fine-tuning recipe with validation

finish-l4-lora-recipeworkflowsetup L32
OriPekelman/toy
What it does

Extract the LoRA fine-tune recipe (L4) — minimal Toy::LLM::Recipes::LoRA wrapping the existing realize_for_mmap + upload_lora_q_init! + training loop

Best for

Orchestrating multi-agent workflows with bounded costs and structured outputs.

Preconditions

Workflow runtime, args with required parameters

Failure modes
  • · Phase timeout or agent failure
  • · Schema validation failure
  • · Missing required args
Trust signals
  • · Comprehensive implementation (>6K LOC)