{"document_type":"rlab.recipe","format_version":1,"provenance":{"asset":null,"runtime":{"image_ref":"docker:ghcr.io/tsilva/rlab/rlab-train@sha256:f358d35bce7a87c17cbf0e642523c2daa78ef4d904895f867208c1cc25e4e323","packages":{"stable-baselines3":"2.8.0","stable-retro-turbo":"1.0.1.post5","supermariobrosnes-turbo":"0.2.5"}},"source_commit":"044dcc3d2a978fb1a77524822238c4c7814c002d","source_files":[]},"recipe":{"description":"Seed 1 runs the Level4-1 transfer of the successful Level1-1 B55 low-KL late-decay recipe, created to test whether that policy recipe transfers without changing PPO or reward hyperparameters while overriding only Level4-1 identity and reporting metadata.","environment":{"env_id":"supermariobrosnes-turbo:SuperMarioBros-Nes-v0","preprocessing":{"frame_skip":4,"frame_stack":4,"max_pool_frames":false,"obs_copy":"safe_view","obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_grayscale":true,"obs_resize":[84,84],"obs_resize_algorithm":"area","pipeline":"supermariobrosnes_turbo_native_vec_env","policy_observation_layout":"channel_first","sticky_action_prob":0.0},"provider_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"schema_version":2,"state":"Level4-1","task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}}},"environment_hash":"sha256:fa0f3dd473fac528938a2cc2130dfe2d516ef164946b6545f2262ff6777c4bef","eval":{"action_sampling":"stochastic","environment":{"env_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"num_threads":4,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"env_provider":"supermariobrosnes-turbo","frame_skip":4,"game":"SuperMarioBros-Nes-v0","max_pool_frames":false,"obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_resize_algorithm":"area","observation_size":84,"state":"Level4-1","sticky_action_prob":0.0,"task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}}},"episodes":100,"max_steps":4500,"n_envs":20,"seed":10007,"seed_protocol":"vector-lane-v1"},"goal":{"eval":{"acceptance":[{"metric":"eval/full/outcome/success/rate/min","operator":">=","threshold":1.0}],"environment":{"env_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"num_threads":4,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"env_provider":"supermariobrosnes-turbo","frame_skip":4,"game":"SuperMarioBros-Nes-v0","max_pool_frames":false,"obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_resize_algorithm":"area","observation_size":84,"state":"Level4-1","sticky_action_prob":0.0,"task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}}},"episodes":100},"goal_id":"Level4-1","objective":{"rank":["max(eval/full/outcome/success/rate/min)","max(eval/full/outcome/success/rate/mean)","min(leader/checkpoint/steps_to_goal)","max(eval/full/episode/return/mean)"]},"release":{"huggingface":{}},"title":"Sample-efficient SuperMarioBros-NES Level4-1 completion","train":{"checkpoint_eval_backend":"local","checkpoint_freq":500000,"environment":{"env_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"num_threads":4,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"env_provider":"supermariobrosnes-turbo","frame_skip":4,"game":"SuperMarioBros-Nes-v0","max_pool_frames":false,"obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_resize_algorithm":"area","observation_size":84,"state":"Level4-1","sticky_action_prob":0.0,"task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}}},"stop_on_acceptance":false}},"recipe_id":"base","schema_version":2,"tags":["goal_id:Level4-1","recipe_id:base","env_id:SuperMarioBros-Nes-v0","level_id:Level4-1"],"train":{"backend":{"config":{"adam_eps":1e-08,"advantage_normalization":"auto","batch_size":512,"clip_range":0.15,"clip_range_vf":null,"device":"auto","ent_coef":0.01,"ent_coef_final":0.0001,"ent_coef_schedule_timesteps":4000000,"gae_lambda":1,"gamma":0.9,"learning_rate":0.00015,"learning_rate_final":0.0001,"learning_rate_schedule_timesteps":4000000,"n_epochs":10,"n_steps":512,"normalize_advantage":false,"policy_net_arch":"","resume":null,"target_kl":0.16,"value_net_arch":"","vf_coef":1},"id":"sb3.ppo"}},"train_config":{"checkpoint_eval_backend":"local","checkpoint_eval_environment":{"env_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"num_threads":4,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"env_provider":"supermariobrosnes-turbo","frame_skip":4,"game":"SuperMarioBros-Nes-v0","max_pool_frames":false,"obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_resize_algorithm":"area","observation_size":84,"state":"Level4-1","sticky_action_prob":0.0,"task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}}},"checkpoint_eval_n_envs":20,"checkpoint_eval_seed":10007,"checkpoint_eval_seed_protocol":"vector-lane-v1","checkpoint_freq":500000,"early_stop":[{"metric":"train/outcome/success/window_100/rate/min","operator":">","threshold":0.99}],"env_args":{"frame_stack":4,"info":"data","info_filter":"all","inttype":"stable","noop_reset_max":0,"num_threads":4,"obs_copy":"safe_view","obs_grayscale":true,"obs_layout":"chw","obs_type":"image","players":1,"record":false,"render_mode":"rgb_array","reward_clip":false,"rom_path":null,"scenario":"scenario","use_restricted_actions":"filtered"},"env_provider":"supermariobrosnes-turbo","frame_skip":4,"game":"SuperMarioBros-Nes-v0","max_pool_frames":false,"n_envs":16,"obs_crop":[32,0,0,0],"obs_crop_fill":0,"obs_crop_mode":"remove","obs_resize_algorithm":"area","observation_size":84,"post_train_eval_episodes":100,"post_train_eval_max_steps":0,"post_train_eval_stochastic":true,"seed":1,"state":"Level4-1","sticky_action_prob":0.0,"stop_on_acceptance":false,"task":{"action":{"set":"simple"},"events":{"level_change":{"operation":"change","signal":"level"},"life_loss":{"operation":"decrease","signal":"lives"}},"id":"mario","reward":{"clip_rewards":false,"completion_reward":0.0,"death_penalty":25.0,"progress_reward_cap":30.0,"progress_reward_scale":1.0,"reward_mode":"score","reward_scale":10.0,"score_progress_clipped":false,"terminal_reward":50.0,"time_penalty":0.0,"use_native_reward":false},"signals":{"level":["levelHi","levelLo"],"lives":"lives","score":"score","x":["xscrollHi","xscrollLo"]},"termination":{"failure":["life_loss"],"max_episode_steps":4500,"success":["level_change"]}},"timesteps":50000000,"training_backend":{"config":{"adam_eps":1e-08,"advantage_normalization":"auto","batch_size":512,"clip_range":0.15,"clip_range_vf":null,"device":"auto","ent_coef":0.01,"ent_coef_final":0.0001,"ent_coef_schedule_timesteps":4000000,"gae_lambda":1,"gamma":0.9,"learning_rate":0.00015,"learning_rate_final":0.0001,"learning_rate_schedule_timesteps":4000000,"n_epochs":10,"n_steps":512,"normalize_advantage":false,"policy_net_arch":"","resume":null,"target_kl":0.16,"value_net_arch":"","vf_coef":1},"id":"sb3.ppo"}}}}