Train a TAMER agent

Based on TAMER by Knox & Stone (2008+)

⚠

Tetris Feedback Tip

Give feedback for the prior piece placement, not the piece currently in motion. Feedback for the current piece will be misinterpreted by TAMER.

⏱

Feedback Timing Tip

Once you toggle training on, try to give feedback (/ or z) as quickly as possible after the action you’re rating. That said, the system does expect small delays and handles them automatically — just do your best!

🎮

Human Control Mode

Controls

Press / for +reward, z for -reward
Tetris tip: Give feedback for the prior piece placement. Do not give feedback for the piece currently in motion, which TAMER will misinterpret.
Training: OFF (Space to toggle)
Mode: Human Control

Status

Episode 0
Steps 0
+Rewards / -Rewards 0 / 0
Speed 1.3 steps/sec

Predicted feedback

Keyboard Shortcuts

Feedback
/ Good (+1)
z Bad (-1)
Space Toggle training
Simulation
1 Start / Pause
2 Single step
+/- Speed
Human Control (press m to toggle)
← / j Left
→ / l Right
↑ / i Up
↓ / k Down/None
(All 4 directions for Loop Maze/Robot Arm)
Actions are sticky — the last key you pressed keeps repeating.