EMO SWARM v2|RUN LIBRARY

Run Library / External Visual

Melting Pot `clean_up`

Externer Benchmark-Visual-Pfad fuer `BenchMARL + Melting Pot`, bewusst getrennt von Live-Training und im Library-Scope nur als offline/inspection entry.

Back To Library

viewer ready

role bridge ready

role bridge emo ready

emo proto slot

trained slot pending

thumbnail only (no motion)

earthlink

task clean_up

Current Winner

all_appetite_pull_080_loyalty_gate_124 | score 4603.00

family role_bridge_emo | source clean_up_role_bridge_emo_gate_repl_j_20260406_164247_j

updated 6.4.2026, 16:42:47

Current Winner Preview

Current Melting Pot clean_up role bridge EMO winner thumbnail

Captured: 4/6/2026, 4:55:47 PM

Next Benchmark Track

Track 2 / territory | substrate TERRITORY__OPEN

status our policy scenario refresh | focal-vs-all closeout capture synced | Track-2 preview now points to a fresh scenario capture from the resumed focal-vs-all closeout (`territory_open1_focal_vs_all_resume_20260410_224214`, trial 3 policy-state). Capture was rendered with 4 sampled episodes and 220 steps for direct visual inspection in the library.

viewer official reference rollout | our policy live

updated 11.4.2026, 00:53:00

Melting Pot territory official reference thumbnail

Our Policy

Captured: 4/11/2026, 12:53:42 AM

Captured: 4/7/2026, 6:10:15 PM

Structural Anchor

Current Melting Pot clean_up role bridge thumbnail

Captured: 4/6/2026, 10:33:05 AM

Current Melting Pot clean_up EMO proto thumbnail

Captured: 4/9/2026, 12:29:18 AM

Trained baseline rollout noch nicht da. Sobald ein Checkpoint existiert, erscheint hier der echte Slot statt eines Platzhalters.

Current Melting Pot clean_up rollout sample thumbnail

Captured: 4/5/2026, 9:43:48 PM

Captured: 4/5/2026, 9:35:02 PM

Der Visual-Pfad verwendet den offiziellen `Melting Pot` `human_players`-Entry-Point nur fuer die nackte Task-Ansicht. Ganz oben steht jetzt zuerst der aktuelle Winner, danach der reward-bearing `role bridge`-Anchor, darunter der aeltere `EMO proto rollout`, dann der `trained baseline rollout` sobald wirklich ein Checkpoint vorliegt, danach das `task rollout sample` und ganz unten die statische Aufgabenkarte. So bleibt klar getrennt, was gerade unser funktionierender Winner ist, was der strukturelle Anchor ist, was nur Emotions-Prototyp war und was bloßer Aufgaben-Kontext ist.

Open Role Bridge Download Role Bridge GIF Open Role Bridge EMO Download Role Bridge EMO GIF Open Track 2 Slot Open Auditor Arena Download Official Reference GIF Download Our Policy GIF Open EMO Viewer Download EMO GIF Open Raw GIF Open Rollout Sample Download Sample GIF Open Static Preview Back To Library

Visual Commands

Role Bridge EMO Sync

cd /Users/mschumann/coding/emo_swarm_v2 && ./scripts/meltingpot_sync_clean_up_role_bridge_emo_rollout.sh --run

Role Bridge Sync

cd /Users/mschumann/coding/emo_swarm_v2 && EMO_MELTING_POT_ROLE_BRIDGE_CLEANERS=3 EMO_MELTING_POT_ROLE_BRIDGE_LAYOUT=center_out ./scripts/meltingpot_sync_clean_up_role_bridge_rollout.sh --run

EMO Proto Sync

cd /Users/mschumann/coding/emo_swarm_v2 && ./scripts/meltingpot_sync_clean_up_emo_rollout.sh --run

Trained Rollout Sync

cd /Users/mschumann/coding/emo_swarm_v2 && ./scripts/meltingpot_sync_clean_up_trained_rollout.sh --run

Rollout Sync

cd /Users/mschumann/coding/emo_swarm_v2 && ./scripts/meltingpot_sync_clean_up_rollout.sh --run

Preview Sync

cd /Users/mschumann/coding/emo_swarm_v2 && ./scripts/meltingpot_sync_clean_up_preview.sh --run

Dry-Run

ssh earthlink "cd /home/mschumann/remote_compute/emo_swarm_v2_melting_pot && ./scripts/meltingpot_visual_clean_up.sh"

Visual Run

ssh earthlink "cd /home/mschumann/remote_compute/emo_swarm_v2_melting_pot && ./scripts/meltingpot_visual_clean_up.sh --run"

Hinweis: Die Library zeigt jetzt vier getrennte Ebenen. `Role Bridge EMO Sync` zieht den aktuellen best Optuna-Slot, `EMO Proto Sync` fuellt den aelteren kleinen Emotionen-Slot, `Trained Rollout Sync` den Checkpoint-Slot, `Rollout Sync` den scripted Episode-Sample und `Preview Sync` die statische Aufgabenkarte.

Baseline / Artifacts

Baseline Run

ssh earthlink "cd /home/mschumann/remote_compute/emo_swarm_v2_melting_pot && ./scripts/meltingpot_benchmarl_baseline.sh --run"

Fuer einen watchable Baseline-Rollout braucht der naechste kleine Run einen End-Checkpoint. Das ist in der Baseline-Script-Default jetzt bereits eingeschaltet.

Output Root

/home/mschumann/remote_compute/emo_swarm_v2_melting_pot/output/melting_pot

Dort landen CSV/JSON-Artefakte fuer den ersten `clean_up`-Baseline-Track.

Why It Lives Here

`Run Library` bleibt laut Contract eine Offline-/Inspection-Oberflaeche. Deshalb liegt der `Melting Pot`-Visual-Pfad hier als sichtbarer Bibliothekseintrag mit Kontext, Commands und Artefaktpfaden, aber ohne versteckte Runtime-Control.