🧠 Coordinator v3 — Optuna Distributed HPO

🧠 Coordinator v3 — Optuna Distributed HPO

🧠 Coordinator v3 — Optuna Distributed HPO

fix: curriculum+classweight+metalearner+delay: training_core.py

fix: curriculum+classweight+metalearner+delay: training_engine.py

fix: local checkpoint fallback + theory-practice connection: model_registry.py

fix: local checkpoint fallback + theory-practice connection: training_core.py

v2.4: training_core (theory+practice separated): app.py

v2.4: training_core (theory+practice separated): training_core.py

Review+deploy: training_engine.py

Review+deploy: requirements.txt

Review+deploy: app.py

T6+T7 complete: forge_controller.py

T6+T7 complete: hypothesis_backtester.py

T5: Full validation layer - WF, regime, benchmark, paper trading

T4-T7: app.py

T4-T7: theory_registry.py

T4-T7: skill_trainers.py

T4-T7: model_registry.py

T4-T7: forge_controller.py

T4-T7: training_engine.py

T2: real Hyperliquid macro data collector

T2: real Hyperliquid macro data collector

v3.0: seed Expert E1-E5 + Meta tasks

v2.2 final: dead code removed, sort fix, import fix

v2.2 audit fixes: BUG-2 FORCE_SKIP_HPO, BUG-4 lock race condition

v2.2: worker heartbeat /worker_heartbeat + /workers dashboard

v2.1: add Worker-10 to fleet watchdog

v2.1: fix restore distributions + FORCE_SKIP_HPO + per-trial try/except

v2.0: task queue + worker watchdog (5min) + self keepalive (30min)

v2.0: Task Queue — HPO→OOS→SKILL, /next_task unified, backward compat

Dual study: TPE (default) + NSGA-II GA (?study=ga). A/B comparison.

Coordinator fix: persist every trial + pending timeout watchdog (120min) + graceful report

📛 Set title: Aetheris Coordinator 🧠

🐛 Fix ValueError on empty study (best_trial)

🧠 Deploy Coordinator v3 — Optuna Distributed HPO

🦆 Switch to Gradio SQL API

🦆 Add SQL Query Tab (DuckDB)

Deploy Dashboard

Deploy Dashboard

Deploy Dashboard

Deploy Dashboard

Deploy Dashboard

initial commit