diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s2000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s2000.yaml new file mode 100644 index 0000000..23bd571 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s2000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s2000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s2000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s3000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s3000.yaml new file mode 100644 index 0000000..8a64308 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s3000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s3000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s3000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s4000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s4000.yaml new file mode 100644 index 0000000..0bd32e6 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s4000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s4000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s4000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s5000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s5000.yaml new file mode 100644 index 0000000..c6356d1 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s5000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s5000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s5000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s6000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s6000.yaml new file mode 100644 index 0000000..5eac8f9 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s6000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s6000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s6000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s7000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s7000.yaml new file mode 100644 index 0000000..979dd5f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed005_s7000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed005_s7000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed005_s7000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s2000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s2000.yaml new file mode 100644 index 0000000..ae67365 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s2000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s2000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s2000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s3000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s3000.yaml new file mode 100644 index 0000000..379b03e --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s3000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s3000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s3000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s4000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s4000.yaml new file mode 100644 index 0000000..a6f4740 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s4000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s4000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s4000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s5000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s5000.yaml new file mode 100644 index 0000000..439943f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s5000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s5000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s5000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s6000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s6000.yaml new file mode 100644 index 0000000..e795a5f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s6000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s6000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s6000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s7000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s7000.yaml new file mode 100644 index 0000000..178b6ff --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed030_s7000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed030_s7000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed030_s7000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s2000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s2000.yaml new file mode 100644 index 0000000..719da1b --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s2000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s2000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s2000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s3000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s3000.yaml new file mode 100644 index 0000000..87d4de3 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s3000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s3000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s3000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s4000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s4000.yaml new file mode 100644 index 0000000..da042e4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s4000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s4000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s4000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s5000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s5000.yaml new file mode 100644 index 0000000..8d95d9f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s5000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s5000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s5000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s6000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s6000.yaml new file mode 100644 index 0000000..45ae71f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s6000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s6000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s6000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s7000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s7000.yaml new file mode 100644 index 0000000..075a256 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed060_s7000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed060_s7000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed060_s7000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s2000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s2000.yaml new file mode 100644 index 0000000..354b7da --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s2000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s2000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s2000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s3000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s3000.yaml new file mode 100644 index 0000000..6c4aafd --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s3000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s3000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s3000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s4000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s4000.yaml new file mode 100644 index 0000000..5680b96 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s4000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s4000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s4000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s5000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s5000.yaml new file mode 100644 index 0000000..3b7787a --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s5000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s5000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s5000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s6000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s6000.yaml new file mode 100644 index 0000000..29b9dc9 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s6000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s6000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s6000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s7000.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s7000.yaml new file mode 100644 index 0000000..d5eb0b0 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_mixed090_s7000.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_mixed090_s7000 + seed: 42 + output_dir: runs/research_v1/eval_axis2_mixed090_s7000 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed000_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed000_s42.yaml new file mode 100644 index 0000000..9089dd7 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed000_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed000_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed000_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed010_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed010_s42.yaml new file mode 100644 index 0000000..a39f740 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed010_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed010_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed010_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed020_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed020_s42.yaml new file mode 100644 index 0000000..11911c0 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed020_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed020_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed020_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed030_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed030_s42.yaml new file mode 100644 index 0000000..eebc81a --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed030_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed030_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed030_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed040_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed040_s42.yaml new file mode 100644 index 0000000..ac06008 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed040_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed040_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed040_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed050_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed050_s42.yaml new file mode 100644 index 0000000..32091a8 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed050_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed050_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed050_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed060_final_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed060_final_s42.yaml new file mode 100644 index 0000000..d34acdb --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed060_final_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed060_final_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed060_final_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed070_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed070_s42.yaml new file mode 100644 index 0000000..5fe2c83 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed070_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed070_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed070_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed080_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed080_s42.yaml new file mode 100644 index 0000000..f8986ad --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed080_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed080_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed080_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed090_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed090_s42.yaml new file mode 100644 index 0000000..f9269d5 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed090_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed090_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed090_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed095_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed095_s42.yaml new file mode 100644 index 0000000..3861b03 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed095_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed095_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed095_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed099_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed099_s42.yaml new file mode 100644 index 0000000..4debb3d --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed099_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed099_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed099_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed100_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed100_s42.yaml new file mode 100644 index 0000000..e813fd4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_mixed100_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_mixed100_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_mixed100_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_uniform_s42.yaml b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_uniform_s42.yaml new file mode 100644 index 0000000..6766949 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_eval/eval_axis2_sampler_uniform_s42.yaml @@ -0,0 +1,25 @@ +experiment: + name: eval_axis2_sampler_uniform_s42 + seed: 42 + output_dir: runs/research_v1/eval_axis2_sampler_uniform_s42 +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 200 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +algorithm: + checkpoint_path: null +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +eval: + scenario_set: heldout_random + n_episodes: 100 + deterministic: true diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s2000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s2000.yaml new file mode 100644 index 0000000..8e4e331 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s2000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s2000 + mode: fasb_ppo + seed: 2000 + output_dir: runs/research_v1/axis2_mixed005_s2000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s3000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s3000.yaml new file mode 100644 index 0000000..e7fef04 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s3000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s3000 + mode: fasb_ppo + seed: 3000 + output_dir: runs/research_v1/axis2_mixed005_s3000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 3000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 3000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s4000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s4000.yaml new file mode 100644 index 0000000..b3bf465 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s4000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s4000 + mode: fasb_ppo + seed: 4000 + output_dir: runs/research_v1/axis2_mixed005_s4000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 4000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 4000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s5000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s5000.yaml new file mode 100644 index 0000000..ab3e3af --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s5000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s5000 + mode: fasb_ppo + seed: 5000 + output_dir: runs/research_v1/axis2_mixed005_s5000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 5000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s6000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s6000.yaml new file mode 100644 index 0000000..cecba5d --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s6000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s6000 + mode: fasb_ppo + seed: 6000 + output_dir: runs/research_v1/axis2_mixed005_s6000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 6000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 6000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s7000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s7000.yaml new file mode 100644 index 0000000..01fa967 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed005_s7000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed005_s7000 + mode: fasb_ppo + seed: 7000 + output_dir: runs/research_v1/axis2_mixed005_s7000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 7000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 7000 + num_scenarios: 500 + failure_ratio: 0.05 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s2000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s2000.yaml new file mode 100644 index 0000000..a54b619 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s2000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s2000 + mode: fasb_ppo + seed: 2000 + output_dir: runs/research_v1/axis2_mixed030_s2000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s3000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s3000.yaml new file mode 100644 index 0000000..50604df --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s3000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s3000 + mode: fasb_ppo + seed: 3000 + output_dir: runs/research_v1/axis2_mixed030_s3000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 3000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 3000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s4000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s4000.yaml new file mode 100644 index 0000000..b530443 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s4000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s4000 + mode: fasb_ppo + seed: 4000 + output_dir: runs/research_v1/axis2_mixed030_s4000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 4000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 4000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s5000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s5000.yaml new file mode 100644 index 0000000..8f767b9 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s5000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s5000 + mode: fasb_ppo + seed: 5000 + output_dir: runs/research_v1/axis2_mixed030_s5000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 5000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s6000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s6000.yaml new file mode 100644 index 0000000..bd2891c --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s6000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s6000 + mode: fasb_ppo + seed: 6000 + output_dir: runs/research_v1/axis2_mixed030_s6000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 6000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 6000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s7000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s7000.yaml new file mode 100644 index 0000000..8fe9033 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed030_s7000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed030_s7000 + mode: fasb_ppo + seed: 7000 + output_dir: runs/research_v1/axis2_mixed030_s7000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 7000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 7000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s2000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s2000.yaml new file mode 100644 index 0000000..35fefc7 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s2000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s2000 + mode: fasb_ppo + seed: 2000 + output_dir: runs/research_v1/axis2_mixed060_s2000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s3000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s3000.yaml new file mode 100644 index 0000000..c12e73f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s3000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s3000 + mode: fasb_ppo + seed: 3000 + output_dir: runs/research_v1/axis2_mixed060_s3000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 3000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 3000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s4000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s4000.yaml new file mode 100644 index 0000000..083e061 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s4000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s4000 + mode: fasb_ppo + seed: 4000 + output_dir: runs/research_v1/axis2_mixed060_s4000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 4000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 4000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s5000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s5000.yaml new file mode 100644 index 0000000..e3d3eef --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s5000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s5000 + mode: fasb_ppo + seed: 5000 + output_dir: runs/research_v1/axis2_mixed060_s5000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 5000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s6000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s6000.yaml new file mode 100644 index 0000000..8d01089 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s6000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s6000 + mode: fasb_ppo + seed: 6000 + output_dir: runs/research_v1/axis2_mixed060_s6000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 6000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 6000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s7000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s7000.yaml new file mode 100644 index 0000000..06c15f1 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed060_s7000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed060_s7000 + mode: fasb_ppo + seed: 7000 + output_dir: runs/research_v1/axis2_mixed060_s7000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 7000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 7000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s2000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s2000.yaml new file mode 100644 index 0000000..868544c --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s2000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s2000 + mode: fasb_ppo + seed: 2000 + output_dir: runs/research_v1/axis2_mixed090_s2000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s3000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s3000.yaml new file mode 100644 index 0000000..8b2d253 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s3000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s3000 + mode: fasb_ppo + seed: 3000 + output_dir: runs/research_v1/axis2_mixed090_s3000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 3000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 3000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s4000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s4000.yaml new file mode 100644 index 0000000..355f8a0 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s4000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s4000 + mode: fasb_ppo + seed: 4000 + output_dir: runs/research_v1/axis2_mixed090_s4000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 4000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 4000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s5000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s5000.yaml new file mode 100644 index 0000000..173c25b --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s5000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s5000 + mode: fasb_ppo + seed: 5000 + output_dir: runs/research_v1/axis2_mixed090_s5000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 5000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 5000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s6000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s6000.yaml new file mode 100644 index 0000000..31667e1 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s6000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s6000 + mode: fasb_ppo + seed: 6000 + output_dir: runs/research_v1/axis2_mixed090_s6000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 6000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 6000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s7000.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s7000.yaml new file mode 100644 index 0000000..10275b4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_mixed090_s7000.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_mixed090_s7000 + mode: fasb_ppo + seed: 7000 + output_dir: runs/research_v1/axis2_mixed090_s7000 + save_resolved_config: true +mode: fasb_ppo +metadrive: + env_class: metadrive.envs.MetaDriveEnv + config: + start_seed: 7000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 7000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed000_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed000_s42.yaml new file mode 100644 index 0000000..b874c58 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed000_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed000_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed000_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.0 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed010_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed010_s42.yaml new file mode 100644 index 0000000..291d811 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed010_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed010_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed010_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.1 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed020_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed020_s42.yaml new file mode 100644 index 0000000..11e3852 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed020_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed020_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed020_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.2 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed030_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed030_s42.yaml new file mode 100644 index 0000000..13d2c92 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed030_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed030_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed030_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.3 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed040_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed040_s42.yaml new file mode 100644 index 0000000..c133d04 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed040_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed040_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed040_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.4 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed050_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed050_s42.yaml new file mode 100644 index 0000000..70149b4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed050_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed050_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed050_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.5 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed060_final_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed060_final_s42.yaml new file mode 100644 index 0000000..9e3675e --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed060_final_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed060_final_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed060_final_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.6 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed070_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed070_s42.yaml new file mode 100644 index 0000000..766769f --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed070_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed070_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed070_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.7 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed080_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed080_s42.yaml new file mode 100644 index 0000000..7438c03 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed080_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed080_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed080_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.8 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed090_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed090_s42.yaml new file mode 100644 index 0000000..1387cb4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed090_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed090_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed090_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.9 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed095_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed095_s42.yaml new file mode 100644 index 0000000..e86ed00 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed095_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed095_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed095_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.95 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed099_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed099_s42.yaml new file mode 100644 index 0000000..9a99d5d --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed099_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed099_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed099_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 0.99 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed100_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed100_s42.yaml new file mode 100644 index 0000000..dccf3f4 --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_mixed100_s42.yaml @@ -0,0 +1,77 @@ +experiment: + name: axis2_sampler_mixed100_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_mixed100_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + metadrive: null + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.MixedFailureSampler + start_seed: 2000 + num_scenarios: 500 + failure_ratio: 1.0 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/resolved_train/axis2_sampler_uniform_s42.yaml b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_uniform_s42.yaml new file mode 100644 index 0000000..ee0fc6a --- /dev/null +++ b/research/research_v1/axis2/configs/resolved_train/axis2_sampler_uniform_s42.yaml @@ -0,0 +1,75 @@ +experiment: + name: axis2_sampler_uniform_s42 + mode: fasb_ppo + seed: 42 + output_dir: runs/research_v1/axis2_sampler_uniform_s42 + save_resolved_config: true +mode: fasb_ppo +metadrive: + config: + start_seed: 2000 + num_scenarios: 500 + traffic_density: 0.1 + random_traffic: true + use_render: false + horizon: 500 + log_level: 50 + crash_vehicle_done: false + crash_object_done: false + out_of_road_done: false +vec_env: + type: dummy + n_envs: 1 + start_method: forkserver +algorithm: + backend: sb3 + name: PPO + policy: MlpPolicy + checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip + params: + learning_rate: 3.0e-05 + n_steps: 128 + batch_size: 64 + n_epochs: 10 + gamma: 0.99 + gae_lambda: 0.95 + clip_range: 0.2 + ent_coef: 0.0 + vf_coef: 0.5 + max_grad_norm: 0.5 + verbose: 1 + device: cpu + policy_kwargs: + net_arch: + - 256 + - 256 +training: + total_timesteps: 300000 + save_every_steps: 100000 + eval_every_steps: 0 +failure_buffer: + path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl + max_size: 5000 +cost_function: + _target_: fasb.plugins.cost.DefaultDrivingCost +failure_scorer: + _target_: fasb.plugins.failure_scorer.DefaultFailureScorer +failure_classifier: + _target_: fasb.plugins.failure_classifier.DefaultFailureClassifier +safety_budget: + _target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget + d_min: 0.1 + d_max: 0.3 + timeout_budget: 0.3 +penalty_scheduler: + _target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler + lambda_min: 0.0 + lambda_max: 0.25 +sampler: + _target_: fasb.plugins.sampler.UniformFailureSampler + start_seed: 2000 + num_scenarios: 500 + alpha: 0.7 + max_too_hard_ratio: 0.15 +eval: + n_episodes: 100 diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed005.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed005.yaml new file mode 100644 index 0000000..906dbbc --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed005.yaml @@ -0,0 +1,14 @@ +experiment: {name: fasb_stable_ratio005_penalty025_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/stabilization/fasb_stable_ratio005_penalty025_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: 0.00003, n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 300000, save_every_steps: 100000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl, max_size: 5000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.10, d_max: 0.30, timeout_budget: 0.30} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.0, lambda_max: 0.25} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 0.05, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed030.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed030.yaml new file mode 100644 index 0000000..1135826 --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed030.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_mixed030_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_mixed030_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: "linear:3.0e-4", n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 100000, save_every_steps: 25000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore/buffers/failure_buffer.jsonl, max_size: 1000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.02, d_max: 0.10, timeout_budget: 0.07} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.1, lambda_max: 5.0} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 0.3, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060.yaml new file mode 100644 index 0000000..2627a3d --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_mixed060_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_mixed060_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: "linear:3.0e-4", n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 100000, save_every_steps: 25000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore/buffers/failure_buffer.jsonl, max_size: 1000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.02, d_max: 0.10, timeout_budget: 0.07} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.1, lambda_max: 5.0} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 0.6, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060_final.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060_final.yaml new file mode 100644 index 0000000..4c9bbd3 --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed060_final.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_mixed060_final_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_mixed060_final_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: 0.00003, n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 300000, save_every_steps: 100000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl, max_size: 5000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.10, d_max: 0.30, timeout_budget: 0.30} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.0, lambda_max: 0.25} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 0.6, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed090.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed090.yaml new file mode 100644 index 0000000..d0d777e --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed090.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_mixed090_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_mixed090_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: "linear:3.0e-4", n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 100000, save_every_steps: 25000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore/buffers/failure_buffer.jsonl, max_size: 1000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.02, d_max: 0.10, timeout_budget: 0.07} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.1, lambda_max: 5.0} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 0.9, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_mixed100.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed100.yaml new file mode 100644 index 0000000..3e7e1b5 --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_mixed100.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_mixed100_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_mixed100_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: "linear:3.0e-4", n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 100000, save_every_steps: 25000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore/buffers/failure_buffer.jsonl, max_size: 1000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.02, d_max: 0.10, timeout_budget: 0.07} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.1, lambda_max: 5.0} +sampler: {_target_: fasb.plugins.sampler.MixedFailureSampler, start_seed: 2000, num_scenarios: 500, failure_ratio: 1.0, alpha: 0.7, max_too_hard_ratio: 0.15} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/configs/templates/axis2_sampler_uniform.yaml b/research/research_v1/axis2/configs/templates/axis2_sampler_uniform.yaml new file mode 100644 index 0000000..2e57e76 --- /dev/null +++ b/research/research_v1/axis2/configs/templates/axis2_sampler_uniform.yaml @@ -0,0 +1,14 @@ +experiment: {name: axis2_sampler_uniform_s42, mode: fasb_ppo, seed: 42, output_dir: runs/research_v1/axis2_sampler_uniform_s42, save_resolved_config: true} +mode: fasb_ppo +metadrive: {env_class: metadrive.envs.MetaDriveEnv, config: {start_seed: 2000, num_scenarios: 500, traffic_density: 0.1, random_traffic: true, use_render: false, horizon: 500, log_level: 50, crash_vehicle_done: false, crash_object_done: false, out_of_road_done: false}} +vec_env: {type: dummy, n_envs: 1, start_method: forkserver} +algorithm: {backend: sb3, name: PPO, policy: MlpPolicy, checkpoint_path: runs/research_v1/base_pretrain_s42/checkpoints/final.zip, params: {learning_rate: "linear:3.0e-4", n_steps: 128, batch_size: 64, n_epochs: 10, gamma: 0.99, gae_lambda: 0.95, clip_range: 0.2, ent_coef: 0.0, vf_coef: 0.5, max_grad_norm: 0.5, verbose: 1, device: cpu, policy_kwargs: {net_arch: [256, 256]}}} +training: {total_timesteps: 100000, save_every_steps: 25000, eval_every_steps: 0} +failure_buffer: {path: runs/research_v1/base_explore/buffers/failure_buffer.jsonl, max_size: 1000} +cost_function: {_target_: fasb.plugins.cost.DefaultDrivingCost} +failure_scorer: {_target_: fasb.plugins.failure_scorer.DefaultFailureScorer} +failure_classifier: {_target_: fasb.plugins.failure_classifier.DefaultFailureClassifier} +safety_budget: {_target_: fasb.plugins.safety_budget.AdaptiveSafetyBudget, d_min: 0.02, d_max: 0.10, timeout_budget: 0.07} +penalty_scheduler: {_target_: fasb.plugins.penalty_scheduler.RiskPenaltyScheduler, lambda_min: 0.1, lambda_max: 5.0} +sampler: {_target_: fasb.plugins.sampler.UniformSampler, start_seed: 2000, num_scenarios: 500} +eval: {n_episodes: 100} diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s2000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s2000_heldout_random.csv new file mode 100644 index 0000000..79a0870 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s2000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.48,0.28,0.49,0.52,0.7066629388176271,-739.5305745162278,-739.5305745162278,260.5,0.0,418.14,-0.55 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s3000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s3000_heldout_random.csv new file mode 100644 index 0000000..977b31b --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s3000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.03,0.13,0.83,0.97,0.6365712250516019,-1687.281736032087,-1687.281736032087,416.01,0.0,495.84,-1.415 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s4000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s4000_heldout_random.csv new file mode 100644 index 0000000..5f0a015 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s4000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.48,0.41,0.35,0.52,0.6936915366792357,-824.0402656986887,-824.0402656986887,262.38,0.0,433.93,-0.54 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s5000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s5000_heldout_random.csv new file mode 100644 index 0000000..05c4b74 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s5000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.45,0.23,0.47,0.55,0.6892048283666706,-673.8237927328566,-673.8237927328566,244.81,0.0,422.39,-0.525 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s6000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s6000_heldout_random.csv new file mode 100644 index 0000000..61c1c6d --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s6000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.46,0.2,0.37,0.54,0.6905250431867175,-725.9684762934507,-725.9684762934507,244.12,0.0,424.17,-0.38 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s7000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s7000_heldout_random.csv new file mode 100644 index 0000000..a76d322 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed005_s7000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.54,0.31,0.36,0.46,0.759461826012154,-695.5406312422863,-695.5406312422863,235.33,0.0,421.26,-0.36 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s2000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s2000_heldout_random.csv new file mode 100644 index 0000000..7a71411 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s2000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.47,0.28,0.44,0.53,0.6978385433350917,-678.2611185633389,-678.2611185633389,239.82,0.0,417.78,-0.5150000000000001 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s3000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s3000_heldout_random.csv new file mode 100644 index 0000000..c783da9 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s3000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.38,0.21,0.44,0.62,0.6540797899642422,-731.287891325107,-731.287891325107,257.77,0.0,422.08,-0.5800000000000001 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s4000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s4000_heldout_random.csv new file mode 100644 index 0000000..1c0567f --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s4000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.48,0.2,0.46,0.52,0.7025671742923127,-683.7937344590678,-683.7937344590678,245.95,0.0,423.93,-0.44 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s5000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s5000_heldout_random.csv new file mode 100644 index 0000000..9a49a07 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s5000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.36,0.24,0.5,0.64,0.6461602085552245,-743.1626915042176,-743.1626915042176,258.64,0.0,434.59,-0.7 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s6000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s6000_heldout_random.csv new file mode 100644 index 0000000..6139bb0 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s6000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.55,0.37,0.39,0.45,0.7419667270799026,-669.3037516183962,-669.3037516183962,240.48,0.0,411.71,-0.4349999999999999 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s7000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s7000_heldout_random.csv new file mode 100644 index 0000000..d2f8a79 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed030_s7000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.0,0.0,0.0,1.0,0.0145519181983123,0.201336443430691,0.201336443430691,0.0,0.0,500.0,-0.5 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s2000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s2000_heldout_random.csv new file mode 100644 index 0000000..8b8fdf0 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s2000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.35,0.18,0.59,0.65,0.6297815623310021,-841.2031631480338,-841.2031631480338,277.59,0.0,435.21,-0.745 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s3000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s3000_heldout_random.csv new file mode 100644 index 0000000..6c32c54 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s3000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.39,0.18,0.46,0.61,0.659454570965486,-782.535332998438,-782.535332998438,261.31,0.0,435.68,-0.5549999999999999 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s4000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s4000_heldout_random.csv new file mode 100644 index 0000000..be77d59 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s4000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.47,0.41,0.46,0.53,0.7246234326492498,-727.5348380024511,-727.5348380024511,250.52,0.0,424.36,-0.665 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s5000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s5000_heldout_random.csv new file mode 100644 index 0000000..701809e --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s5000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.47,0.33,0.42,0.53,0.6868272247103674,-714.3171086278159,-714.3171086278159,245.36,0.0,426.44,-0.545 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s6000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s6000_heldout_random.csv new file mode 100644 index 0000000..0bd2e3d --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s6000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.43,0.25,0.49,0.58,0.690117515691638,-695.8891576970805,-695.8891576970805,253.47,0.0,424.73,-0.6 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s7000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s7000_heldout_random.csv new file mode 100644 index 0000000..686713f --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed060_s7000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.42,0.16,0.44,0.58,0.6618710321026859,-766.4502476648471,-766.4502476648471,267.51,0.0,427.27,-0.47 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s2000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s2000_heldout_random.csv new file mode 100644 index 0000000..d1121f6 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s2000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.42,0.25,0.48,0.58,0.7003562557136677,-697.7429937521977,-697.7429937521977,253.22,0.0,421.42,-0.6 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s3000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s3000_heldout_random.csv new file mode 100644 index 0000000..31d3db7 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s3000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.46,0.23,0.37,0.54,0.6936422935248217,-623.9836426340675,-623.9836426340675,233.17,0.0,419.89,-0.41 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s4000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s4000_heldout_random.csv new file mode 100644 index 0000000..cf04194 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s4000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.45,0.26,0.4,0.55,0.6644866569242734,-708.6731081327355,-708.6731081327355,239.46,0.0,414.85,-0.485 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s5000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s5000_heldout_random.csv new file mode 100644 index 0000000..4a164d6 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s5000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.41,0.2,0.49,0.59,0.6593525444596314,-763.0033888325256,-763.0033888325256,254.04,0.0,429.65,-0.575 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s6000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s6000_heldout_random.csv new file mode 100644 index 0000000..f29698e --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s6000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.39,0.2,0.44,0.61,0.656586863162787,-800.8696573616239,-800.8696573616239,266.09,0.0,429.88,-0.5549999999999999 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s7000_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s7000_heldout_random.csv new file mode 100644 index 0000000..91d9015 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_mixed090_s7000_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.44,0.24,0.51,0.56,0.7202393162743849,-629.9968468110172,-629.9968468110172,238.95,0.0,422.69,-0.5900000000000001 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed000_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed000_s42_heldout_random.csv new file mode 100644 index 0000000..35b3c32 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed000_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.37,0.15,0.52,0.64,0.6292836273247241,-851.1962263209364,-851.1962263209364,271.25,0.0,433.89,-0.6200000000000001 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed010_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed010_s42_heldout_random.csv new file mode 100644 index 0000000..7c56bf3 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed010_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.42,0.18,0.43,0.58,0.7165654727731726,-735.4927292230267,-735.4927292230267,262.91,0.0,428.7,-0.48 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed020_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed020_s42_heldout_random.csv new file mode 100644 index 0000000..f140d58 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed020_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.52,0.26,0.42,0.48,0.7347442224560331,-624.4479850099999,-624.4479850099999,220.91,0.0,429.76,-0.3999999999999999 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed030_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed030_s42_heldout_random.csv new file mode 100644 index 0000000..56178d3 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed030_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.47,0.37,0.45,0.53,0.6775630028696646,-770.8565883184264,-770.8565883184264,255.41,0.0,429.2,-0.615 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed040_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed040_s42_heldout_random.csv new file mode 100644 index 0000000..4151076 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed040_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.46,0.26,0.46,0.54,0.6737452199413194,-707.5653101566407,-707.5653101566407,242.91,0.0,414.06,-0.53 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed050_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed050_s42_heldout_random.csv new file mode 100644 index 0000000..27d13c0 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed050_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.49,0.21,0.48,0.51,0.7223570841155599,-669.8456865358968,-669.8456865358968,246.71,0.0,405.9,-0.4549999999999999 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed060_final_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed060_final_s42_heldout_random.csv new file mode 100644 index 0000000..fd29888 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed060_final_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.42,0.31,0.55,0.58,0.6648399432957663,-740.8202435676453,-740.8202435676453,261.0,0.0,421.79,-0.73 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed070_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed070_s42_heldout_random.csv new file mode 100644 index 0000000..0d293ef --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed070_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.45,0.24,0.43,0.55,0.6655551456331337,-706.681869426326,-706.681869426326,252.09,0.0,427.14,-0.495 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed080_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed080_s42_heldout_random.csv new file mode 100644 index 0000000..f9aa530 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed080_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.47,0.26,0.41,0.54,0.6945954260516823,-660.8659046247266,-660.8659046247266,244.82,0.0,417.49,-0.47 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed090_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed090_s42_heldout_random.csv new file mode 100644 index 0000000..03065d2 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed090_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.41,0.2,0.5,0.59,0.6405156240153242,-806.5244893892319,-806.5244893892319,274.97,0.0,432.09,-0.585 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed095_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed095_s42_heldout_random.csv new file mode 100644 index 0000000..e13f28e --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed095_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.44,0.17,0.41,0.56,0.6682853590806817,-652.5512946428923,-652.5512946428923,240.71,0.0,417.8,-0.42 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed099_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed099_s42_heldout_random.csv new file mode 100644 index 0000000..86939ae --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed099_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.48,0.26,0.47,0.52,0.7030364695841829,-656.0966138291578,-656.0966138291578,236.1,0.0,420.75,-0.51 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed100_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed100_s42_heldout_random.csv new file mode 100644 index 0000000..06c4d24 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_mixed100_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.5,0.28,0.41,0.5,0.7155517678613296,-813.5187156639124,-813.5187156639124,256.39,0.0,435.94,-0.44 diff --git a/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_uniform_s42_heldout_random.csv b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_uniform_s42_heldout_random.csv new file mode 100644 index 0000000..70ccd26 --- /dev/null +++ b/research/research_v1/axis2/results/final_eval/eval_axis2_sampler_uniform_s42_heldout_random.csv @@ -0,0 +1,2 @@ +n_episodes,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_reward_mean,episode_modified_reward_mean,episode_cost_mean,cost_violation_rate,avg_episode_length,safety_efficiency_score +100.0,0.53,0.21,0.36,0.47,0.6976322573087572,-663.5564626749714,-663.5564626749714,243.17,0.0,414.14,-0.2749999999999999 diff --git a/research/research_v1/axis2/results/summary/multiseed_axis2_per_seed.csv b/research/research_v1/axis2/results/summary/multiseed_axis2_per_seed.csv new file mode 100644 index 0000000..af20642 --- /dev/null +++ b/research/research_v1/axis2/results/summary/multiseed_axis2_per_seed.csv @@ -0,0 +1,25 @@ +seed,method,eval_run,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_cost_mean,safety_efficiency_score +2000,mixed005,eval_axis2_mixed005_s2000,0.48,0.28,0.49,0.52,0.7066629388176271,260.5,-0.55 +3000,mixed005,eval_axis2_mixed005_s3000,0.03,0.13,0.83,0.97,0.6365712250516019,416.01,-1.415 +4000,mixed005,eval_axis2_mixed005_s4000,0.48,0.41,0.35,0.52,0.6936915366792357,262.38,-0.54 +5000,mixed005,eval_axis2_mixed005_s5000,0.45,0.23,0.47,0.55,0.6892048283666706,244.81,-0.525 +6000,mixed005,eval_axis2_mixed005_s6000,0.46,0.2,0.37,0.54,0.6905250431867175,244.12,-0.38 +7000,mixed005,eval_axis2_mixed005_s7000,0.54,0.31,0.36,0.46,0.759461826012154,235.33,-0.36 +2000,mixed030,eval_axis2_mixed030_s2000,0.47,0.28,0.44,0.53,0.6978385433350917,239.82,-0.5150000000000001 +3000,mixed030,eval_axis2_mixed030_s3000,0.38,0.21,0.44,0.62,0.6540797899642422,257.77,-0.5800000000000001 +4000,mixed030,eval_axis2_mixed030_s4000,0.48,0.2,0.46,0.52,0.7025671742923127,245.95,-0.44 +5000,mixed030,eval_axis2_mixed030_s5000,0.36,0.24,0.5,0.64,0.6461602085552245,258.64,-0.7 +6000,mixed030,eval_axis2_mixed030_s6000,0.55,0.37,0.39,0.45,0.7419667270799026,240.48,-0.4349999999999999 +7000,mixed030,eval_axis2_mixed030_s7000,0.0,0.0,0.0,1.0,0.0145519181983123,0.0,-0.5 +2000,mixed060,eval_axis2_mixed060_s2000,0.35,0.18,0.59,0.65,0.6297815623310021,277.59,-0.745 +3000,mixed060,eval_axis2_mixed060_s3000,0.39,0.18,0.46,0.61,0.659454570965486,261.31,-0.5549999999999999 +4000,mixed060,eval_axis2_mixed060_s4000,0.47,0.41,0.46,0.53,0.7246234326492498,250.52,-0.665 +5000,mixed060,eval_axis2_mixed060_s5000,0.47,0.33,0.42,0.53,0.6868272247103674,245.36,-0.545 +6000,mixed060,eval_axis2_mixed060_s6000,0.43,0.25,0.49,0.58,0.690117515691638,253.47,-0.6 +7000,mixed060,eval_axis2_mixed060_s7000,0.42,0.16,0.44,0.58,0.6618710321026859,267.51,-0.47 +2000,mixed090,eval_axis2_mixed090_s2000,0.42,0.25,0.48,0.58,0.7003562557136677,253.22,-0.6 +3000,mixed090,eval_axis2_mixed090_s3000,0.46,0.23,0.37,0.54,0.6936422935248217,233.17,-0.41 +4000,mixed090,eval_axis2_mixed090_s4000,0.45,0.26,0.4,0.55,0.6644866569242734,239.46,-0.485 +5000,mixed090,eval_axis2_mixed090_s5000,0.41,0.2,0.49,0.59,0.6593525444596314,254.04,-0.575 +6000,mixed090,eval_axis2_mixed090_s6000,0.39,0.2,0.44,0.61,0.656586863162787,266.09,-0.5549999999999999 +7000,mixed090,eval_axis2_mixed090_s7000,0.44,0.24,0.51,0.56,0.7202393162743849,238.95,-0.5900000000000001 diff --git a/research/research_v1/axis2/results/summary/multiseed_axis2_summary.csv b/research/research_v1/axis2/results/summary/multiseed_axis2_summary.csv new file mode 100644 index 0000000..5a81edc --- /dev/null +++ b/research/research_v1/axis2/results/summary/multiseed_axis2_summary.csv @@ -0,0 +1,9 @@ +block,block_trains,method,n_seeds,success_rate_mean,success_rate_std,collision_rate_mean,collision_rate_std,offroad_rate_mean,offroad_rate_std,timeout_rate_mean,timeout_rate_std,route_completion_mean_mean,route_completion_mean_std,episode_cost_mean_mean,episode_cost_mean_std,safety_efficiency_score_mean,safety_efficiency_score_std +6_train_first3,6,mixed005,3,0.33,0.2598076211353316,0.2733333333333334,0.140118997046558,0.5566666666666666,0.24684678108764824,0.6699999999999999,0.25980762113533157,0.6789752335161549,0.03729127604085676,312.9633333333333,89.2459816088844,-0.835,0.5023196193659969 +6_train_first3,6,mixed030,3,0.44333333333333336,0.05507570547286101,0.22999999999999998,0.04358898943540675,0.4466666666666667,0.011547005383792525,0.5566666666666666,0.05507570547286101,0.6848285025305488,0.026733920046383138,247.84666666666666,9.12406890226796,-0.5116666666666667,0.07005949852327904 +6_train_first3,6,mixed060,3,0.4033333333333333,0.06110100926607786,0.25666666666666665,0.1327905619136139,0.5033333333333333,0.07505553499465131,0.5966666666666667,0.06110100926607786,0.6712865219819126,0.048515372718419526,263.14,13.62746858370987,-0.6549999999999999,0.09539392014169461 +6_train_first3,6,mixed090,3,0.44333333333333336,0.020816659994661348,0.24666666666666667,0.015275252316519465,0.4166666666666667,0.056862407030773256,0.5566666666666668,0.020816659994661285,0.6861617353875875,0.019068982454798995,241.95000000000002,10.254301536428509,-0.49833333333333335,0.09569918146637062 +12_train_all6,12,mixed005,6,0.4066666666666667,0.18715412543320192,0.26,0.09674709297958257,0.4783333333333333,0.18225440095280734,0.5933333333333333,0.18715412543320184,0.6960195663523344,0.0393811049422864,277.19166666666666,68.79423447256802,-0.6283333333333333,0.39424188852361525 +12_train_all6,12,mixed030,6,0.37333333333333335,0.19572088970436108,0.21666666666666665,0.12274635093014646,0.37166666666666665,0.1855172947912584,0.6266666666666667,0.19572088970436105,0.5761940602375144,0.27736310706937406,207.10999999999999,101.79231326578642,-0.5283333333333334,0.09963265863494093 +12_train_all6,12,mixed060,6,0.42166666666666663,0.046654760385909884,0.25166666666666665,0.09988326519826364,0.4766666666666666,0.06022181221672647,0.5800000000000001,0.046475800154488996,0.6754458897417382,0.03252215230916063,259.29333333333335,11.9278575891342,-0.5966666666666667,0.09709102258533829 +12_train_all6,12,mixed090,6,0.4283333333333333,0.026394443859772212,0.22999999999999998,0.025298221281347028,0.4483333333333334,0.0549241901776136,0.5716666666666667,0.026394443859772177,0.682443988343261,0.02607301144720482,247.48833333333334,12.362607195355944,-0.5358333333333333,0.0739876115756325 diff --git a/research/research_v1/axis2/results/summary/single_seed_axis2_sweep.csv b/research/research_v1/axis2/results/summary/single_seed_axis2_sweep.csv new file mode 100644 index 0000000..0121ddf --- /dev/null +++ b/research/research_v1/axis2/results/summary/single_seed_axis2_sweep.csv @@ -0,0 +1,15 @@ +failure_ratio,method,success_rate,collision_rate,offroad_rate,timeout_rate,route_completion_mean,episode_cost_mean,safety_efficiency_score +0.0,axis2_sampler_mixed000,0.37,0.15,0.52,0.64,0.6292836273247241,271.25,-0.6200000000000001 +0.1,axis2_sampler_mixed010,0.42,0.18,0.43,0.58,0.7165654727731726,262.91,-0.48 +0.2,axis2_sampler_mixed020,0.52,0.26,0.42,0.48,0.7347442224560331,220.91,-0.3999999999999999 +0.3,axis2_sampler_mixed030,0.47,0.37,0.45,0.53,0.6775630028696646,255.41,-0.615 +0.4,axis2_sampler_mixed040,0.46,0.26,0.46,0.54,0.6737452199413194,242.91,-0.53 +0.5,axis2_sampler_mixed050,0.49,0.21,0.48,0.51,0.7223570841155599,246.71,-0.4549999999999999 +0.6,axis2_sampler_mixed060_final,0.42,0.31,0.55,0.58,0.6648399432957663,261.0,-0.73 +0.7,axis2_sampler_mixed070,0.45,0.24,0.43,0.55,0.6655551456331337,252.09,-0.495 +0.8,axis2_sampler_mixed080,0.47,0.26,0.41,0.54,0.6945954260516823,244.82,-0.47 +0.9,axis2_sampler_mixed090,0.41,0.2,0.5,0.59,0.6405156240153242,274.97,-0.585 +0.95,axis2_sampler_mixed095,0.44,0.17,0.41,0.56,0.6682853590806817,240.71,-0.42 +0.99,axis2_sampler_mixed099,0.48,0.26,0.47,0.52,0.7030364695841829,236.1,-0.51 +1.0,axis2_sampler_mixed100,0.5,0.28,0.41,0.5,0.7155517678613296,256.39,-0.44 +uniform,axis2_sampler_uniform,0.53,0.21,0.36,0.47,0.6976322573087572,243.17,-0.2749999999999999 diff --git a/scripts/generate_axis2_configs.py b/scripts/generate_axis2_configs.py new file mode 100644 index 0000000..0d17b5f --- /dev/null +++ b/scripts/generate_axis2_configs.py @@ -0,0 +1,494 @@ +import shutil +from pathlib import Path +import yaml +import re + +def main(): + base_dir = Path(r"c:\Users\white\Desktop\hw\DRL\hw5\MetaDrive-Experiment-Framework") + research_dir = base_dir / "research" / "research_v1" / "axis2" + configs_dir = research_dir / "configs" + + templates_dir = configs_dir / "templates" + resolved_train_dir = configs_dir / "resolved_train" + resolved_eval_dir = configs_dir / "resolved_eval" + + templates_dir.mkdir(parents=True, exist_ok=True) + resolved_train_dir.mkdir(parents=True, exist_ok=True) + resolved_eval_dir.mkdir(parents=True, exist_ok=True) + + # 1. Copy Template YAMLs + template_files = { + base_dir / "configs" / "research_v1" / "axis2_sampler_mixed060_final.yaml": "axis2_sampler_mixed060_final.yaml", + base_dir / "configs" / "research_v1" / "screening" / "axis2_sampler_mixed030.yaml": "axis2_sampler_mixed030.yaml", + base_dir / "configs" / "research_v1" / "screening" / "axis2_sampler_mixed060.yaml": "axis2_sampler_mixed060.yaml", + base_dir / "configs" / "research_v1" / "screening" / "axis2_sampler_mixed090.yaml": "axis2_sampler_mixed090.yaml", + base_dir / "configs" / "research_v1" / "screening" / "axis2_sampler_mixed100.yaml": "axis2_sampler_mixed100.yaml", + base_dir / "configs" / "research_v1" / "screening" / "axis2_sampler_uniform.yaml": "axis2_sampler_uniform.yaml", + base_dir / "configs" / "research_v1" / "stabilization" / "fasb_stable_ratio005_penalty025.yaml": "axis2_sampler_mixed005.yaml" + } + + for src, dest_name in template_files.items(): + if src.exists(): + shutil.copy2(src, templates_dir / dest_name) + print(f"Copied template: {dest_name}") + else: + print(f"Warning: Template source {src} does not exist.") + + # 2. Generate Resolved Train and Eval YAMLs dynamically based on the multiseed runs + # Methods: mixed005, mixed030, mixed060, mixed090 + # Seeds: 2000, 3000, 4000, 5000, 6000, 7000 + methods_ratios = { + "mixed005": 0.05, + "mixed030": 0.30, + "mixed060": 0.60, + "mixed090": 0.90 + } + seeds = [2000, 3000, 4000, 5000, 6000, 7000] + + for method, ratio in methods_ratios.items(): + for seed in seeds: + eval_run = f"eval_axis2_{method}_s{seed}" + train_run = f"axis2_{method}_s{seed}" + + # --- Resolved Train Config --- + train_config = { + "experiment": { + "name": train_run, + "mode": "fasb_ppo", + "seed": seed, + "output_dir": f"runs/research_v1/{train_run}", + "save_resolved_config": True + }, + "mode": "fasb_ppo", + "metadrive": { + "env_class": "metadrive.envs.MetaDriveEnv", + "config": { + "start_seed": seed, + "num_scenarios": 500, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "vec_env": { + "type": "dummy", + "n_envs": 1, + "start_method": "forkserver" + }, + "algorithm": { + "backend": "sb3", + "name": "PPO", + "policy": "MlpPolicy", + "checkpoint_path": "runs/research_v1/base_pretrain_s42/checkpoints/final.zip", + "params": { + "learning_rate": 3.0e-05, + "n_steps": 128, + "batch_size": 64, + "n_epochs": 10, + "gamma": 0.99, + "gae_lambda": 0.95, + "clip_range": 0.2, + "ent_coef": 0.0, + "vf_coef": 0.5, + "max_grad_norm": 0.5, + "verbose": 1, + "device": "cpu", + "policy_kwargs": { + "net_arch": [256, 256] + } + } + }, + "training": { + "total_timesteps": 300000, + "save_every_steps": 100000, + "eval_every_steps": 0 + }, + "failure_buffer": { + "path": "runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl", + "max_size": 5000 + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "failure_scorer": { + "_target_": "fasb.plugins.failure_scorer.DefaultFailureScorer" + }, + "failure_classifier": { + "_target_": "fasb.plugins.failure_classifier.DefaultFailureClassifier" + }, + "safety_budget": { + "_target_": "fasb.plugins.safety_budget.AdaptiveSafetyBudget", + "d_min": 0.1, + "d_max": 0.3, + "timeout_budget": 0.3 + }, + "penalty_scheduler": { + "_target_": "fasb.plugins.penalty_scheduler.RiskPenaltyScheduler", + "lambda_min": 0.0, + "lambda_max": 0.25 + }, + "sampler": { + "_target_": "fasb.plugins.sampler.MixedFailureSampler", + "start_seed": seed, + "num_scenarios": 500, + "failure_ratio": ratio, + "alpha": 0.7, + "max_too_hard_ratio": 0.15 + }, + "eval": { + "n_episodes": 100 + } + } + + # --- Resolved Eval Config --- + eval_config = { + "experiment": { + "name": eval_run, + "seed": 42, + "output_dir": f"runs/research_v1/{eval_run}" + }, + "metadrive": { + "env_class": "metadrive.envs.MetaDriveEnv", + "config": { + "start_seed": 5000, + "num_scenarios": 200, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "algorithm": { + "checkpoint_path": None + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "eval": { + "scenario_set": "heldout_random", + "n_episodes": 100, + "deterministic": True + } + } + + # Write YAMLs using standard formatting + train_file = resolved_train_dir / f"{train_run}.yaml" + with open(train_file, "w") as f: + yaml.dump(train_config, f, default_flow_style=False, sort_keys=False) + + eval_file = resolved_eval_dir / f"{eval_run}.yaml" + with open(eval_file, "w") as f: + yaml.dump(eval_config, f, default_flow_style=False, sort_keys=False) + + # 3. Write Seed 42 Sweep Resolved Train and Eval YAMLs + # Single-seed sweep: ratio 0.00 to 1.00 (from sampler yaml configs) + sampler_sweep = { + "axis2_sampler_mixed000_s42": 0.00, + "axis2_sampler_mixed010_s42": 0.10, + "axis2_sampler_mixed020_s42": 0.20, + "axis2_sampler_mixed030_s42": 0.30, + "axis2_sampler_mixed040_s42": 0.40, + "axis2_sampler_mixed050_s42": 0.50, + "axis2_sampler_mixed060_final_s42": 0.60, + "axis2_sampler_mixed070_s42": 0.70, + "axis2_sampler_mixed080_s42": 0.80, + "axis2_sampler_mixed090_s42": 0.90, + "axis2_sampler_mixed095_s42": 0.95, + "axis2_sampler_mixed099_s42": 0.99, + "axis2_sampler_mixed100_s42": 1.00, + } + + for train_run, ratio in sampler_sweep.items(): + eval_run = f"eval_{train_run}" + seed = 42 + + # Train config + train_config = { + "experiment": { + "name": train_run, + "mode": "fasb_ppo", + "seed": seed, + "output_dir": f"runs/research_v1/{train_run}", + "save_resolved_config": True + }, + "mode": "fasb_ppo", + "metadrive": { + "metadrive": None, # Will resolve to default env_class + "config": { + "start_seed": 2000, + "num_scenarios": 500, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "vec_env": { + "type": "dummy", + "n_envs": 1, + "start_method": "forkserver" + }, + "algorithm": { + "backend": "sb3", + "name": "PPO", + "policy": "MlpPolicy", + "checkpoint_path": "runs/research_v1/base_pretrain_s42/checkpoints/final.zip", + "params": { + "learning_rate": 3.0e-05, + "n_steps": 128, + "batch_size": 64, + "n_epochs": 10, + "gamma": 0.99, + "gae_lambda": 0.95, + "clip_range": 0.2, + "ent_coef": 0.0, + "vf_coef": 0.5, + "max_grad_norm": 0.5, + "verbose": 1, + "device": "cpu", + "policy_kwargs": { + "net_arch": [256, 256] + } + } + }, + "training": { + "total_timesteps": 300000, + "save_every_steps": 100000, + "eval_every_steps": 0 + }, + "failure_buffer": { + "path": "runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl", + "max_size": 5000 + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "failure_scorer": { + "_target_": "fasb.plugins.failure_scorer.DefaultFailureScorer" + }, + "failure_classifier": { + "_target_": "fasb.plugins.failure_classifier.DefaultFailureClassifier" + }, + "safety_budget": { + "_target_": "fasb.plugins.safety_budget.AdaptiveSafetyBudget", + "d_min": 0.1, + "d_max": 0.3, + "timeout_budget": 0.3 + }, + "penalty_scheduler": { + "_target_": "fasb.plugins.penalty_scheduler.RiskPenaltyScheduler", + "lambda_min": 0.0, + "lambda_max": 0.25 + }, + "sampler": { + "_target_": "fasb.plugins.sampler.MixedFailureSampler", + "start_seed": 2000, + "num_scenarios": 500, + "failure_ratio": ratio, + "alpha": 0.7, + "max_too_hard_ratio": 0.15 + }, + "eval": { + "n_episodes": 100 + } + } + + # Eval config + eval_config = { + "experiment": { + "name": eval_run, + "seed": 42, + "output_dir": f"runs/research_v1/{eval_run}" + }, + "metadrive": { + "env_class": "metadrive.envs.MetaDriveEnv", + "config": { + "start_seed": 5000, + "num_scenarios": 200, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "algorithm": { + "checkpoint_path": None + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "eval": { + "scenario_set": "heldout_random", + "n_episodes": 100, + "deterministic": True + } + } + + # Write YAMLs + train_file = resolved_train_dir / f"{train_run}.yaml" + with open(train_file, "w") as f: + yaml.dump(train_config, f, default_flow_style=False, sort_keys=False) + + eval_file = resolved_eval_dir / f"{eval_run}.yaml" + with open(eval_file, "w") as f: + yaml.dump(eval_config, f, default_flow_style=False, sort_keys=False) + + # 4. Uniform Sampler + train_run = "axis2_sampler_uniform_s42" + eval_run = "eval_axis2_sampler_uniform_s42" + + # Train config (Uniform) + train_config = { + "experiment": { + "name": train_run, + "mode": "fasb_ppo", + "seed": 42, + "output_dir": f"runs/research_v1/{train_run}", + "save_resolved_config": True + }, + "mode": "fasb_ppo", + "metadrive": { + "config": { + "start_seed": 2000, + "num_scenarios": 500, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "vec_env": { + "type": "dummy", + "n_envs": 1, + "start_method": "forkserver" + }, + "algorithm": { + "backend": "sb3", + "name": "PPO", + "policy": "MlpPolicy", + "checkpoint_path": "runs/research_v1/base_pretrain_s42/checkpoints/final.zip", + "params": { + "learning_rate": 3.0e-05, + "n_steps": 128, + "batch_size": 64, + "n_epochs": 10, + "gamma": 0.99, + "gae_lambda": 0.95, + "clip_range": 0.2, + "ent_coef": 0.0, + "vf_coef": 0.5, + "max_grad_norm": 0.5, + "verbose": 1, + "device": "cpu", + "policy_kwargs": { + "net_arch": [256, 256] + } + } + }, + "training": { + "total_timesteps": 300000, + "save_every_steps": 100000, + "eval_every_steps": 0 + }, + "failure_buffer": { + "path": "runs/research_v1/base_explore_large/buffers/failure_buffer.jsonl", + "max_size": 5000 + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "failure_scorer": { + "_target_": "fasb.plugins.failure_scorer.DefaultFailureScorer" + }, + "failure_classifier": { + "_target_": "fasb.plugins.failure_classifier.DefaultFailureClassifier" + }, + "safety_budget": { + "_target_": "fasb.plugins.safety_budget.AdaptiveSafetyBudget", + "d_min": 0.1, + "d_max": 0.3, + "timeout_budget": 0.3 + }, + "penalty_scheduler": { + "_target_": "fasb.plugins.penalty_scheduler.RiskPenaltyScheduler", + "lambda_min": 0.0, + "lambda_max": 0.25 + }, + "sampler": { + "_target_": "fasb.plugins.sampler.UniformFailureSampler", + "start_seed": 2000, + "num_scenarios": 500, + "alpha": 0.7, + "max_too_hard_ratio": 0.15 + }, + "eval": { + "n_episodes": 100 + } + } + + # Write Uniform + with open(resolved_train_dir / f"{train_run}.yaml", "w") as f: + yaml.dump(train_config, f, default_flow_style=False, sort_keys=False) + + # Eval config + eval_config = { + "experiment": { + "name": eval_run, + "seed": 42, + "output_dir": f"runs/research_v1/{eval_run}" + }, + "metadrive": { + "env_class": "metadrive.envs.MetaDriveEnv", + "config": { + "start_seed": 5000, + "num_scenarios": 200, + "traffic_density": 0.1, + "random_traffic": True, + "use_render": False, + "horizon": 500, + "log_level": 50, + "crash_vehicle_done": False, + "crash_object_done": False, + "out_of_road_done": False + } + }, + "algorithm": { + "checkpoint_path": None + }, + "cost_function": { + "_target_": "fasb.plugins.cost.DefaultDrivingCost" + }, + "eval": { + "scenario_set": "heldout_random", + "n_episodes": 100, + "deterministic": True + } + } + with open(resolved_eval_dir / f"{eval_run}.yaml", "w") as f: + yaml.dump(eval_config, f, default_flow_style=False, sort_keys=False) + + print("All template and resolved train/eval configs generated successfully under research/research_v1/axis2/configs/!") + +if __name__ == "__main__": + main() diff --git a/scripts/organize_axis_data.py b/scripts/organize_axis_data.py new file mode 100644 index 0000000..8097558 --- /dev/null +++ b/scripts/organize_axis_data.py @@ -0,0 +1,166 @@ +import pandas as pd +from pathlib import Path +import re + +def main(): + csv_path = Path(r"C:\Users\white\Downloads\summary_main_results.csv") + if not csv_path.exists(): + print(f"Error: {csv_path} does not exist.") + return + + df = pd.read_csv(csv_path) + print(f"Loaded {len(df)} rows from {csv_path}") + + # Standard headers for individual eval CSV files + eval_cols = [ + "n_episodes", "success_rate", "collision_rate", "offroad_rate", + "timeout_rate", "route_completion_mean", "episode_reward_mean", + "episode_modified_reward_mean", "episode_cost_mean", + "cost_violation_rate", "avg_episode_length", "safety_efficiency_score" + ] + + # Create base paths + base_dir = Path(r"c:\Users\white\Desktop\hw\DRL\hw5\MetaDrive-Experiment-Framework\research\research_v1") + + # Process each row + axis2_rows = [] + + for _, row in df.iterrows(): + method_name = str(row["method"]) + eval_run = str(row["eval_run"]) + + # Determine Axis (Only Axis 2 is processed on this branch) + axis = None + if "axis2" in method_name or "axis2" in eval_run: + axis = 2 + else: + continue + + # Extract seed + seed_match = re.search(r'_s(\d+)$', eval_run) + if seed_match: + seed = int(seed_match.group(1)) + elif "eval_base_pretrain" in eval_run: + seed = 42 + else: + seed = 42 + + # Create output directories + axis_dir = base_dir / f"axis{axis}" + final_eval_dir = axis_dir / "results" / "final_eval" + final_eval_dir.mkdir(parents=True, exist_ok=True) + + # Write individual CSV + eval_file = final_eval_dir / f"{eval_run}_heldout_random.csv" + # Select matching columns + row_dict = row.to_dict() + eval_row = {col: row_dict.get(col, 0.0) for col in eval_cols} + pd.DataFrame([eval_row])[eval_cols].to_csv(eval_file, index=False) + # print(f"Wrote: {eval_file}") + + # Store for aggregation + row_dict["seed"] = seed + row_dict["axis"] = axis + if axis == 2: + axis2_rows.append(row_dict) + + print("Finished writing individual eval CSVs.") + + # --- AGGREGATE AXIS 2 --- + if axis2_rows: + axis2_df = pd.DataFrame(axis2_rows) + summary_dir = base_dir / "axis2" / "results" / "summary" + summary_dir.mkdir(parents=True, exist_ok=True) + reports_dir = base_dir / "axis2" / "reports" + reports_dir.mkdir(parents=True, exist_ok=True) + others_dir = base_dir / "axis2" / "others" + others_dir.mkdir(parents=True, exist_ok=True) + + # 1. Multi-seed Variance Analysis (seeds 2000-7000, methods mixed005, mixed030, mixed060, mixed090) + ms_df = axis2_df[axis2_df["seed"].isin([2000, 3000, 4000, 5000, 6000, 7000]) & axis2_df["method"].str.startswith("axis2_mixed")].copy() + + # We need a clean method name (e.g. mixed005, mixed030) + def clean_method(m): + m = m.replace("axis2_", "") + m = re.sub(r'_s\d+$', '', m) + return m + ms_df["method_clean"] = ms_df["method"].apply(clean_method) + + # Sort values for consistency + ms_df = ms_df.sort_values(by=["method_clean", "seed"]) + + # Write multiseed_axis2_per_seed.csv + per_seed_cols = ["seed", "method", "eval_run", "success_rate", "collision_rate", "offroad_rate", "timeout_rate", "route_completion_mean", "episode_cost_mean", "safety_efficiency_score"] + # Use clean method name in per-seed output + per_seed_df = ms_df.copy() + per_seed_df["method"] = per_seed_df["method_clean"] + per_seed_df[per_seed_cols].to_csv(summary_dir / "multiseed_axis2_per_seed.csv", index=False) + print(f"Wrote: {summary_dir / 'multiseed_axis2_per_seed.csv'}") + + # Compute summary stats (3 seeds vs 6 seeds) + summary_rows = [] + metrics = ["success_rate", "collision_rate", "offroad_rate", "timeout_rate", "route_completion_mean", "episode_cost_mean", "safety_efficiency_score"] + + # 6_train_first3 (seeds 2000, 3000, 4000) + first3_df = ms_df[ms_df["seed"].isin([2000, 3000, 4000])] + for method_name, group in first3_df.groupby("method_clean"): + row = { + "block": "6_train_first3", + "block_trains": 6, + "method": method_name, + "n_seeds": len(group) + } + for m in metrics: + row[f"{m}_mean"] = group[m].mean() + row[f"{m}_std"] = group[m].std() + summary_rows.append(row) + + # 12_train_all6 (seeds 2000, 3000, 4000, 5000, 6000, 7000) + for method_name, group in ms_df.groupby("method_clean"): + row = { + "block": "12_train_all6", + "block_trains": 12, + "method": method_name, + "n_seeds": len(group) + } + for m in metrics: + row[f"{m}_mean"] = group[m].mean() + row[f"{m}_std"] = group[m].std() + summary_rows.append(row) + + summary_df = pd.DataFrame(summary_rows) + # Order columns to match Axis 1 + summary_cols = ["block", "block_trains", "method", "n_seeds"] + for m in metrics: + summary_cols.extend([f"{m}_mean", f"{m}_std"]) + summary_df[summary_cols].to_csv(summary_dir / "multiseed_axis2_summary.csv", index=False) + print(f"Wrote: {summary_dir / 'multiseed_axis2_summary.csv'}") + + # 2. Single-seed Sweep (seed 42, methods starts with axis2_sampler_) + ss_df = axis2_df[axis2_df["seed"] == 42].copy() + + def extract_ratio(m): + if "uniform" in m: + return "uniform" + match = re.search(r'mixed(\d+)', m) + if match: + val = int(match.group(1)) + if val == 95: + return 0.95 + if val == 99: + return 0.99 + return val / 100.0 + return "unknown" + + ss_df["failure_ratio"] = ss_df["method"].apply(extract_ratio) + ss_df = ss_df.sort_values(by="failure_ratio", key=lambda x: x.map(lambda v: v if isinstance(v, float) else (1.01 if v == "uniform" else -1))) + + sweep_cols = ["failure_ratio", "method", "success_rate", "collision_rate", "offroad_rate", "timeout_rate", "route_completion_mean", "episode_cost_mean", "safety_efficiency_score"] + ss_df[sweep_cols].to_csv(summary_dir / "single_seed_axis2_sweep.csv", index=False) + print(f"Wrote: {summary_dir / 'single_seed_axis2_sweep.csv'}") + + + print("All data structured and aggregated successfully!") + +if __name__ == "__main__": + main()