39 lines
1.5 KiB
YAML
39 lines
1.5 KiB
YAML
# Configuration for the evaluation script (eval.py)
|
|
|
|
# List of game names to process
|
|
# Example: ["twenty_forty_eight", "sokoban"]
|
|
game_list: ["sokoban", "candy_crush", "tetris", "twenty_forty_eight", "super_mario_bros"]
|
|
|
|
# List of model names/prefixes to process
|
|
# Example: ["claude-3-5-sonnet-20241022", "gpt-4o"]
|
|
model_list: ["gemini-2.0-flash"]
|
|
|
|
# Force update existing entries in model_perf_rank.json (boolean)
|
|
force_model_perf_rank: false
|
|
|
|
# Force update existing entries in game_perf.json (boolean)
|
|
force_game_perf: true
|
|
|
|
# Generate bar plot (boolean)
|
|
generate_bar_plot: true
|
|
|
|
# Generate radar chart (boolean)
|
|
generate_radar_chart: true
|
|
|
|
# Generate video replays for supported games (boolean)
|
|
# Currently supports "twenty_forty_eight" median score replay from game_perf.json step_infos.
|
|
generate_replays: true
|
|
|
|
# Polynomial Model Configuration
|
|
run_polynomial_model: false # Set to true to run the polynomial model analysis using its internal defaults
|
|
|
|
# File paths (can be overridden if necessary, defaults are usually fine)
|
|
# model_perf_rank_file: "eval/perf/model_perf_rank.json"
|
|
# detailed_game_perf_file: "eval/perf/game_perf.json"
|
|
# other_task_rank_file: "eval/perf/other_task_rank.json" # For polynomial model
|
|
# polynomial_model_results_file: "eval/perf/polynomial_model.json" # For polynomial model
|
|
# plot_output_dir: "eval/perf/plots"
|
|
# video_output_base_dir: "eval/perf/video"
|
|
# model_colors_file: "eval/assets/model_colors.json"
|
|
# game_specific_config_file_path: "eval/configs/game_eval_config.yaml"
|