1
Fork 0

feat(evaluate): evaluation scripts

This commit is contained in:
Tibo De Peuter 2026-05-07 23:43:39 +02:00
parent b88a7ac660
commit 56afc7ec5d
Signed by: tdpeuter
SSH key fingerprint: SHA256:u/h/LVoqKF1Iz02uOyxe6hcjmoZASCGV2HM0TG9ZMoU
6 changed files with 390 additions and 126 deletions

View file

@ -9,12 +9,9 @@ eval_seed: 0
# Cross-model comparison settings
# We use 10 episodes to get a more robust average for the final poster results.
comparison_base_seed: 0
comparison_num_episodes: 10
comparison_output_csv: "metrics/poster_comparison.csv"
comparison_num_episodes: 2
comparison_output_csv: "runs/evaluation/comparison.csv"
# Paths to the .cleanrl_model files to be compared (relative to workspace root).
# These are placeholders; replace with actual trained model paths for the poster.
comparison_models:
- "experiments/poster/centralized.cleanrl_model"
- "experiments/poster/decentralized.cleanrl_model"
- "experiments/poster/decentralized_amputated.cleanrl_model"
- "runs/input-space-2-arms/2026-05-02/08-14-58/final_model.flax"