Blog

A Coding Implementation on Microsoft SkillOpt for Instrumented Prompt Optimization, Skill Evolution Analysis, and Baseline Comparison

k = RUN_KNOBS train_out = run_cli(}", f"train.batch_size={k}", f"gradient.minibatch_size={k}", f"gradient.merge_batch_size={k}", f"gradient.analyst_workers={k}", f"optimizer.learning_rate={k}", f"optimizer.lr_scheduler={k}", "optimizer.use_slow_update=true", "optimizer.use_meta_skill=true", f"env.workers={k}", f"env.limit={k}"], "TRAIN (rollout->reflect->aggregate->select->update->gate; slow-update +...

Pin It on Pinterest