diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..eda7a80b51cc8cede176d77f2a5838e1abce00df --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ec8f71129e88c590a6a6a873527c722e4384ed0839cd2ba70d0c00a02577712a +size 16154320 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7c889e30603e5397dfaf08749bcfdf6301dc1c6b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs10_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs10 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs10/seed_10", + "wandb_unique_id": "deadly-corridor-fs10-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..82b6ee04bdfc9b16e9e83de5af0e7bb18b759aa5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a02a80129d36d625629e199825eb93d2d011ad8495745863e27befdde3e737f9 +size 16154320 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d58b44aff03e0fd67631371d7d58eb34b875e15 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs10_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs10 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs10/seed_11", + "wandb_unique_id": "deadly-corridor-fs10-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..7506a9487e6c10ba8c8532ee084a1e27104a9d76 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cfa6dbb065291bca85c70a27281fe5213e38def8bdb240d75c867a073d5a6dde +size 16154320 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0438610aa5c926c12cabff416cc6277561b30b8b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs10_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs10 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs10/seed_12", + "wandb_unique_id": "deadly-corridor-fs10-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e8669511099d7294ec96eeffe590e30d3095c2fc --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:29605984642c87a4d0e807074e2f0fe9d3e02f557932945925619525eb68f748 +size 16154953 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8a6e34830829301dfc7125ff8143c76460094171 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs10_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs10 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs10/seed_13", + "wandb_unique_id": "deadly-corridor-fs10-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8f20ba00de242dc6fbd2d7f828ea669478fa4ecf --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0112a01cc9461ff455982651e797006ab12e22b3c178fadb7501eaca7c4bb586 +size 16154320 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5efb0cee380b2d7d99218efe06931afb5e0cfe10 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs10:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs10_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs10 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs10/seed_14", + "wandb_unique_id": "deadly-corridor-fs10-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..e9c1f180441c99faa2cd7d30243afa0a57c99bb7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a06c6977e4a7da6e4cf705258d92f600215eba8e6c2bef6d39cd1af05ff10f4b +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c9af39781516c1109c7bf7b2bab0f87601b056a4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs12_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs12 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs12/seed_10", + "wandb_unique_id": "deadly-corridor-fs12-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..09d83d2d23462b2cd9c6e31b27c540afdfdf25ad --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0ddc234f92784fef3ccad3a3f64d8ac8855a0ea299de4da5149badca5e9ec40b +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d4f6ea9baf782325da12f869f50ce0be108d8a97 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs12_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs12 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs12/seed_11", + "wandb_unique_id": "deadly-corridor-fs12-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..fadb1a658d50716d554d59feeca2d7d0f920d085 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:20f363a39b66146808560995595465b7a1e4b9a424d988fe9741ce1cca81931d +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f76b447c94f8e34b360cc40b6191020057853984 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs12_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs12 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs12/seed_12", + "wandb_unique_id": "deadly-corridor-fs12-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..727b9443e47a6b76b5853d400c4ce6fb0497f728 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:605f91872196a05cd9debfc0488955a120c17f72b55c449af97aaf1a984eb31e +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..255e7936e45dc090f7d0d048e673aed96f8f9a1e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs12_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs12 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs12/seed_13", + "wandb_unique_id": "deadly-corridor-fs12-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..75474c65eb4200a438b2331451140beccffcfe11 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a4f3e45c09a1e621713278d7f6b2f132b6faad58549a317c281f274cbb468840 +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..06db69b898930aca1975800754c12bbd1725db7e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs12:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs12_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs12 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs12/seed_14", + "wandb_unique_id": "deadly-corridor-fs12-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..a57c49362c11367d27f1c9e0e61755da5ebafec0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23a03c39dd7bc5ccfbc29cfe37040a0ef68a64148310017313f9cafb7e0558f8 +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bb3787265fbc71b0ce254ab24aeffe8ee709efe4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs14_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs14 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs14/seed_10", + "wandb_unique_id": "deadly-corridor-fs14-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e06d6e3379b290cf695005bb43afb483988e6164 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19dca664c0267c3bb292d5584c6165893ed6210e1de2b171dc96e91b11485520 +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e30d0b4564ed8b033a625cf5b98bcb42dd75dca1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs14_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs14 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs14/seed_11", + "wandb_unique_id": "deadly-corridor-fs14-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..2ddcdb366f581308c9fe57e77eb97fd8e4ea8ff6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e166f307d72aacb0b690aea3283142a7c69a13fe560d1344f048340f2f2851f6 +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ed38966aa95e5694a5f96868ff0052c8a59925f5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs14_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs14 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs14/seed_12", + "wandb_unique_id": "deadly-corridor-fs14-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..21fac5bb49c07cec804f846939087c818434e96d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73dd071c395344b13e4ccaae4ac439ef05012e344d2e3ff0f72b8d2fc233d26a +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b84a6114d51d45236e12dc8f897e995685f0c4f4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs14_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs14 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs14/seed_13", + "wandb_unique_id": "deadly-corridor-fs14-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..d609b9190527686341e8590a53bd7066f4d0428f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b454e2f1402c4766658e1b636a6ca0654371b07de284113de92068fa79b30fdb +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..69fcca3addf8cbae1bdb2cfebeb43ac553db9f19 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs14:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs14_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs14 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs14_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs14", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs14_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs14/seed_14", + "wandb_unique_id": "deadly-corridor-fs14-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..52e9a07a4c08847d2241d03dcee880fa0e0015b9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7e968bde655391dec6b99d4ed4963a9ebcc54e6055f92e7bf8abf011bc47ade6 +size 18955984 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ad6d7133cd589525f9ce9709fee2fbb32a52dacb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs16_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs16 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs16/seed_10", + "wandb_unique_id": "deadly-corridor-fs16-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..952a1ea75a8180c6c35feabe100f04ac2c94b5b9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:decc9a1e23880ea0c2a7fd259fa0581d21bc47a09fee668e7e94a6da64991749 +size 18955984 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f7329013be20f6e6e89ff0fc28c5412ca712caf3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs16_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs16 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs16/seed_11", + "wandb_unique_id": "deadly-corridor-fs16-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e61f23c8bc019b069e53a26ec95aa982449dd87a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2f39f6ec34845d010adacdccd44ebe9edb07125288dab3d3f059a9346fda545 +size 18955984 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6263e7f45ca066a6544c6f166c399d0b08d7353f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs16_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs16 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs16/seed_12", + "wandb_unique_id": "deadly-corridor-fs16-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..34d15c55030fe9732f25427c852ab451e02cbfe2 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3188561d018e71610e5b622f4fa12a43f3d418e29aeede3f16903fc6d01880d1 +size 18955984 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0bd348c67867b1b2eb4774f23f743b90bf56524 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs16_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs16 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs16/seed_13", + "wandb_unique_id": "deadly-corridor-fs16-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..731e7105b22bcc74292914aae507e5d37d7b7993 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:466efa09fd60e489e3f1040e7e5f4ac0efc364817bb01c285feb136ea9e5f593 +size 18955984 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4d6481a986007b82326ea175aebe5380674906b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs16:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs16_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs16 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs16_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs16", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs16_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs16/seed_14", + "wandb_unique_id": "deadly-corridor-fs16-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..284baa0266871bec28985d1187eed7b236e70b7f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e71a5a21fb853427f796a3ed20881b6f4b779b194c979713a60237b36acd14b6 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..817611081c68acfa86761436cbf453d65f809228 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs1_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs1 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs1/seed_10", + "wandb_unique_id": "deadly-corridor-fs1-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8da15aa7c77f9df3cab98b5b7fe1a49238f3ac9d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0c90d9a0315318d2c6292fb6ffb54a45565ae8a4581b22ff0a691401de96b538 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ad95158287807d68c77d807feba7e596e862a824 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs1_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs1 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs1/seed_11", + "wandb_unique_id": "deadly-corridor-fs1-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..2adf2d0d8a6f7af76b51ba4d272a5db92a4dbe6d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ee59ee2f63868924a065e88171819ef4dba85cfd83b267728410f53b39336927 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..824dd468ec013a5b2e8fce82ff573a135953aaf7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs1_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs1 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs1/seed_12", + "wandb_unique_id": "deadly-corridor-fs1-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a381ffd3a8c867f665f8dfde3e252e0525194824 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5e48ed781f044c7a2cadaecdf3bdccedfd1c7674f4b54a2edcde0898f89e2fb2 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3f8c2aa77c525bd79e9e70640f20f3762de5fd52 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs1_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs1 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs1/seed_13", + "wandb_unique_id": "deadly-corridor-fs1-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..9ffcb9fff564cc622686a67888cd2fce04c26a72 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4d4132fecedb74e8670b2199b1f68c536bf76c9298a6396e1e7df94de53427f4 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e0d1c917dbf0358177ad660ff70af30493d9672 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs1:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs1_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs1 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs1", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs1_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs1/seed_14", + "wandb_unique_id": "deadly-corridor-fs1-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..6dce70be2be2f356820e48eb129b99ed2fc48ecd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c8d7f7449c5a6407c5292ac6f07e0fc8beca7b1a1a2d0273ef883e5dbbe1f64a +size 12418768 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4dc6335c0e806900ef10c33fae89386fad4b8929 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs2_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs2 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs2/seed_10", + "wandb_unique_id": "deadly-corridor-fs2-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..3315908da412d60c78d71db3372612f052e5695e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ceb2c0a6105cd9f7f5e1682ca7dfbc20d332ba2e6f1bd64854dd251033b5b7ba +size 12418768 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..19d6fa46cc826b84e39af4313c31187968b814de --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs2_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs2 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs2/seed_11", + "wandb_unique_id": "deadly-corridor-fs2-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a168eeac36f7051ac47cc3782a8b4a524725c175 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9860392c15da60fb2cb23e8201a0e5093fc94eaf86b2866423e7a614b3ff514c +size 12418768 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0b57041fa4cd7454681e94088088feea7785f221 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs2_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs2 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs2/seed_12", + "wandb_unique_id": "deadly-corridor-fs2-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..356db74bbf51efdd2b17454f7263cbe6b87cf0b5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57177e4a186abfd4fc357cb56f499b61ea942ec8e71234daabc07acff96b8127 +size 12418768 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ca9320a39335c60a4ba29ad3974c13e05788333b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs2_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs2 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs2/seed_13", + "wandb_unique_id": "deadly-corridor-fs2-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..954c6f38f912e949e6794eea70d3f6ab71a17de9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9803646d6b98d11a832368678d75ef3708221f6c748f338793ee2c67a7292d3d +size 12418768 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..55730e797db83d352b20322dedccea71b00742e7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs2:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs2_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs2 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs2", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs2_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs2/seed_14", + "wandb_unique_id": "deadly-corridor-fs2-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1440e24bdff030cf73004f318c7ba84ba8b07b2 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae4c6af65cd2ec6f205e61a5c9af9a05cc34d96448257ddcd6dd02ff1fbe4395 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..14fab9bc5099160fae2268f6fe2d8ba05e007208 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs4_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs4 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs4/seed_10", + "wandb_unique_id": "deadly-corridor-fs4-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..974e865f1197c49d802f62710e8adbf70a6a2044 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:dff4dd3c2b052d70db2dc14187f08b07ac2d55dd3bb3237b298f4727b493b2fc +size 13352656 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5200ea0fae9544901d43af5140652ab8d4c676e3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs4_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs4 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs4/seed_11", + "wandb_unique_id": "deadly-corridor-fs4-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..416c559dd5950b08eeafc2a08217fd5c47035616 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e9a33cc0127f49ef94ac731e33660d3c9481e9415133e0d426ac4a124b0cd215 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7a597e1054a04df1ca2fb7c066750431af1297d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs4_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs4 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs4/seed_12", + "wandb_unique_id": "deadly-corridor-fs4-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e4fd11056f14da7d948ddfb46c4d7a2fd11dafca --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e10b907cc80bb6dd0112ee7a41de87b173dbe93bf4ff2d2bcac4e811b900e74a +size 13352656 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..879b6f639e0a173ff436c6c9fa3f2298e736f940 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs4_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs4 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs4/seed_13", + "wandb_unique_id": "deadly-corridor-fs4-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..00915787940a86b89c70f92a9821cbbf6ef9e3c7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f0221bd21a36b95048d0dec216690fa14ab48a97db6db3786833fa696752176 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fe5886287a2f1e9fcaa868c8533c8add1e542650 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs4:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs4_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs4 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs4/seed_14", + "wandb_unique_id": "deadly-corridor-fs4-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..d7d81eff12415bc94f5931a257ccb6a586d08489 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:453f13c120f50bebb131fae5e898446c3848833d0cb7c4aaa1f9c822c80bdd20 +size 14286544 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..50691ee7e0cfbd3ad693f0505891e14baed57ece --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs6_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs6 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs6/seed_10", + "wandb_unique_id": "deadly-corridor-fs6-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..b7709b0a1652fe6062836a77b1e9e0613a6d3783 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad47432f24d90522d7bdb519c73545f36198089e4543f24246f7ec86290af09c +size 14286544 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..74731eb34a0db886dc65ef9be18acb79a8dc0167 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs6_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs6 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs6/seed_11", + "wandb_unique_id": "deadly-corridor-fs6-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..49f63a4ab67822861b1e56b9469ea021d66156db --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f0178a3861747eed67757b413ba2f45acd98a821c3b7069fa7e396cc76b91d43 +size 14286544 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..81d98184e809bdfde1b9d86434c0254562b80deb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs6_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs6 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs6/seed_12", + "wandb_unique_id": "deadly-corridor-fs6-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..621cd39e627f50df7326dfd4b6c07a61fbe6aa4e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bbe5639deaec1a54d10746d35f54ff2870fcfa7874d0402cbc68797c32cb7d3b +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b26ddf2809c43c6fcbea3fb6bac5bd336b5184c9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs6_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs6 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs6/seed_13", + "wandb_unique_id": "deadly-corridor-fs6-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..b3e066a420f0c8d00ed2921bacb64cc2c254f4f7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8ffc4f7ca94f6986841c10b6a3232191072ae319aec9f55707aac6517a2cb6cb +size 14286544 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ca381aaf54312566d364a5c3f08b2f7361a984d4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs6:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs6 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs6/seed_14", + "wandb_unique_id": "deadly-corridor-fs6-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..32b3a1ade92d32645f1e2142e6b0ecdfeea09003 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75df22d1d3a25d7b573278c4740d1a8112eacff5cbcc40c7a73a4b7d86eae5cb +size 15220432 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c38c0b6edd2de809ad1db308ec074eb2e66ae8a5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs8_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs8 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs8/seed_10", + "wandb_unique_id": "deadly-corridor-fs8-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..fa06a38fa15591271700b7b500033454a14e97e3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:251f06d6364d9c4a12706af87e8cc211e2863d762d3834ec49e695e1f5dcef90 +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1c78cd02699fdbf2356fd3c848b4a583c05635bf --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs8_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs8 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs8/seed_11", + "wandb_unique_id": "deadly-corridor-fs8-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..b52c9564f8daccca0ef77e3ffecf179654bd0a0b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:85d0a6ae0e575aaf7079c5c927c3dc71742fcd7af3fed2c0dbc4838d3f5e11fd +size 15220432 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f68e3470013a5743033256904f2a4337217b99b6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs8_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs8 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs8/seed_12", + "wandb_unique_id": "deadly-corridor-fs8-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..91f450a9e7e6124ea1d3c5617822a6b3306e6fbd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c2669306f648c36ab20698997d703f400dc28b697a519925fec2b8f96663dfec +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7163fc9e75d5dc3ae1c2d1af8722afd2b544329e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs8_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs8 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs8/seed_13", + "wandb_unique_id": "deadly-corridor-fs8-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..0650eefd5274c2b939aadc3d683e89dde9b3c5fa --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:19f578d4bdc45a02d4fdd6806fcd61ca92b23c9f6f9c04132714c4b9b84a78c0 +size 15220432 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..78ba9046027d31bc5db65385a87f9df786a57486 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:fixed_l4:fs8:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_fixed_l4_fs8_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack fixed fixed_l4 fs8 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_fixed_l4_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/fixed_l4/fs8/seed_14", + "wandb_unique_id": "deadly-corridor-fs8-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..0e3351f1c7cebaadef022ae29e5adb0ad8f50856 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e58a8835102a0237752ee6cbc93d2d160ae18ed2475219f46a3863483ed748d3 +size 16154953 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bb74e13c6797c7d0730364c1ac50e67f51933d36 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs10 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs10/seed_10", + "wandb_unique_id": "deadly-corridor-fs10-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..c46122c3942fb721316a59072582a16f24dee73f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ea5430c394fb63b396df00d2d17c8ff92f4853ed81cd9e4897c977b21bbb5097 +size 16154953 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9d2be75a301a64e01a4fb233524988ab7cf2a18b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs10 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs10/seed_11", + "wandb_unique_id": "deadly-corridor-fs10-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..c898abc18e373da33e9c79e877e17bccc7a4f954 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:caf1c1cd90893e76e24b8657b6377fe7a4a035b822df5dcaabd9770e33f7f64b +size 16154320 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9fa77c1d8b03a49fc1e55d9344129301ddfb4f21 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs10 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs10/seed_12", + "wandb_unique_id": "deadly-corridor-fs10-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..83be5771000e9c183af7bb4a1adc0a10f731b4a1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6a53ed379077d9b0950bb10c13c118d74a9e5ef413df2b86b39944d4abc9ead +size 16154953 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b2ffbf6bd63f934a79638c203ad348d5e72086e1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs10 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs10/seed_13", + "wandb_unique_id": "deadly-corridor-fs10-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..719d28a9a324cdd33a35a15fb5eb8994dd592cff --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a473d42985808ca749e5d1fdbb1bf9ccb197558ccb9303111ff53035fd049de2 +size 16154953 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..967e2811d2a73000e52d366dc89c5acf693c6aa6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs10:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs10-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs10 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 10 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs10-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs10", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 10, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs10_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs10/seed_14", + "wandb_unique_id": "deadly-corridor-fs10-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..bcf784bb2e6243eff6efe9826644d77d40f5aebe --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f9f85170dadf11983246a79aefb566a2c1621e4dab95178351747c5a31acb73 +size 17088841 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c519b7ce15f2e7b698d02eb406df89347a6fc056 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs12 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs12/seed_10", + "wandb_unique_id": "deadly-corridor-fs12-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a57987bd3a017e636a048db694c95b0ff16d2074 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d1a96a6ddf256f2436468f2175d277c1948dc2bab1ce58d0990dc99272660382 +size 17088841 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9e3f89d1b06958a4a3961e1cec6106780261efdf --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs12 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs12/seed_11", + "wandb_unique_id": "deadly-corridor-fs12-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..5d8314f1eece98ae04f003db537706d7afa42846 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa05c2c39d43a7df4e9b1a46bb70097b960a1a6dcd5141dea66ee8eea0dd1bc0 +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..320b360c57a8079fcc93dbaa5fc08f08c8ce311a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs12 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs12/seed_12", + "wandb_unique_id": "deadly-corridor-fs12-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..b9188f8d197841e75b7a4f63b5e1c28282da4d76 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f70352f679c817ad1310c2334fa1668b97a40649d7ff52e8c3d4491394334a4 +size 17088208 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e1d85a552ee6ad60122056de3bb3d2ac4bc1d926 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs12 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs12/seed_13", + "wandb_unique_id": "deadly-corridor-fs12-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e449b6d793e412cbf9a49e04f5ecb08936bb7946 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b3c164c3df695707c80351ae442ba9d6e05b89d8dfbbfa5599cdb2c734446087 +size 17088841 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39b6f8976d6cd659b11197b2b894a5421307f181 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs12:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs12-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs12 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 12 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs12-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs12", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 12, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs12_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs12/seed_14", + "wandb_unique_id": "deadly-corridor-fs12-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..f0792a09b4d7657fa4e1b097160919b8dc348187 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:520581f3ec2121072444d1f7c25d1b3f6ff4dd1beddd9ce851e760bb3bfe3b78 +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d6095b967f361c093eab3b77e6cd3f70f1668cab --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs14 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs14/seed_10", + "wandb_unique_id": "deadly-corridor-fs14-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..dfc0f7cfc90911ba5e8c2f887494d1a137978874 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:375e86d871ed14c11b8dd45df19eb1da8da874ac62082545e2ae771c39b3c65d +size 18022096 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..52ae6bd4f7d8e138a12a7244f2a0f874d5e08f94 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs14 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs14/seed_11", + "wandb_unique_id": "deadly-corridor-fs14-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..323943e9f2e9bf15c87249e827bb69ed45e4f027 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:eb3e90c60b94b9ca67582a00e693d4b4291aff2a454c913f05162af64425b5c1 +size 18022729 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fb4cfc0e00e70886e03482f262f049c31db6d323 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs14 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs14/seed_12", + "wandb_unique_id": "deadly-corridor-fs14-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8e5f2ca6e465fba8c22ec32834bb5f7086d734dd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a52805bc29a17b925d910e1f0308dde5c7b37e5234bb0e55fe5e220a64247f5 +size 18022729 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0960143cc70ee515ca052c0d6fc8b0331b44f587 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs14 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs14/seed_13", + "wandb_unique_id": "deadly-corridor-fs14-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..4ba4010554b842253790ecdd74d0c07d06619045 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f1fed2971d8f82b51a377f375503bdccdd0d63130ae8af56c687484ccccf4b72 +size 18022729 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..811023f711be78122f7010b667d5b4a1fd76517a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs14:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs14-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs14 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 14 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs14-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs14", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 14, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs14_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs14/seed_14", + "wandb_unique_id": "deadly-corridor-fs14-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..df2418432d0e8e0af24a56e7e68b7b74b0c96947 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a9801b8b2288e719679632bd5960bfd007ee809d05eed9b433571a1d07d63a0 +size 18956617 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6f1418b0a0ea97b047b7d12d3ab14307a7b87482 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs16 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs16/seed_10", + "wandb_unique_id": "deadly-corridor-fs16-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..b5f5df4a2ffd8d8d64e57d5f08034c424f957c44 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a7f9fad14e89d756629cb0bc585e78bfe1b29b53935b53ec64ab29ad99c50b2f +size 18956617 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9161122b6dc0eaef5b4bb57a3dfddbb23ac5c19b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs16 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs16/seed_11", + "wandb_unique_id": "deadly-corridor-fs16-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..c1894de83243850c304eb6264391502d6d0d8e1d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:917b2c8f7c7fedb538d53c44c0586c63c9386b6a443f12ca46f9eec27e39e335 +size 18956617 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..10d537a43263a29f8771c5a86cc48c71afe279fb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs16 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs16/seed_12", + "wandb_unique_id": "deadly-corridor-fs16-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..e5b28626ddd97ad65147a2cf429fc41e1ecc2459 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:562d5eeb87c8f972dedbf76452a448e5a216e493566fe467e795ed472957af3a +size 18956617 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..318774af15a2310ea0b06ca6d130ef4f16deb4ba --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs16 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs16/seed_13", + "wandb_unique_id": "deadly-corridor-fs16-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..714ba42b12452afb9efdb0ab15ae2542661d6774 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92896fd3f483c674242370f1c1c5fcd59f067614556f7d578b265ee0b6633ebd +size 18956617 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6aabbd823e86f9dba907fe19eccdce1fd7a55e27 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs16:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs16-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs16 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 16 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs16-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs16", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 16, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs16_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs16/seed_14", + "wandb_unique_id": "deadly-corridor-fs16-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..0de79939e1956ae60eaddb61cd1cf6e139e8e155 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:55c073f5abe198e0df73cffd7d8d9bb9c3344583019afb7d806da30b13a86f65 +size 11951824 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..92d35544fc0c2d4bc3ce66e5667b1097c1e4bcc1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs1 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs1/seed_10", + "wandb_unique_id": "deadly-corridor-fs1-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..711d0aae0472a28f2d490f0af33f5494e714006a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44e2d137a211c414425eaeeb5b83345f15292ec66ed499b6b41c7b1acc5b1641 +size 11952457 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5f6fac634030386653b9f1a0ef8a411935780398 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs1 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs1/seed_11", + "wandb_unique_id": "deadly-corridor-fs1-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..86ebe65e40d19a65ccebc47a46354bb68012a516 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2ff9414380a6edbdaaa504bea2aa21d5e85ee5ff283f47fe3057b3593d35d1b8 +size 11952457 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e1f469e99f570d7cb0b3811af3164a8b9e121398 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs1 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs1/seed_12", + "wandb_unique_id": "deadly-corridor-fs1-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..980c6f996345bc44e654aea4273a94970ced63bb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:026d78189dc742fb8f445b4e1ba427a4fcacae7986ac9da7104cf20353a45739 +size 11952457 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..409a12555f45f02defc096f9040a7a4e52d933bc --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs1 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs1/seed_13", + "wandb_unique_id": "deadly-corridor-fs1-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..acef6867cbd94592144241cca0efb421a52a4c18 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2775ce579c0efffa7d047b3466b5bda14519314a5133c5d6a17bead6e6c4f90 +size 11952457 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..675d7eecc4140fc33aadf86e2994e0d7a909bdc5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs1:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs1 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 1 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs1", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 1, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs1_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs1/seed_14", + "wandb_unique_id": "deadly-corridor-fs1-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..ea0c50408f695581cafc5a1531508785baac6c17 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e3afe3b44c8fa97e92534d8ceeaec786298e83dc0a88d9ab90b906ef21ee956a +size 12419401 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..93468b0446171a4fe861f69c8d79233262f28eb8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs2 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs2/seed_10", + "wandb_unique_id": "deadly-corridor-fs2-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..1a989e2474447e6f8c00b68f7af244368e387203 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:01325e87ebf1c5b87a3b4d73b59ff5cc6534e21f0722f25ad83379fa1b733cea +size 12419401 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..03212e7be485306b7049abc5f4edf36bca751deb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs2 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs2/seed_11", + "wandb_unique_id": "deadly-corridor-fs2-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..f57756736c6d46ea9ee3707787df23e0d006746e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e4107d566c2189a78b8b90a03035eba2ba6e617d96183cb71a228562d84d696d +size 12419401 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..35123d1143337bd65bc63cda9a9ca286021bd222 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs2 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs2/seed_12", + "wandb_unique_id": "deadly-corridor-fs2-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..f4e5e6bc7e2928f225bdae066ab5c2226497acb9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:446f09f7c24c607b2c35c938be1d7d539f8cce8d270d656d62f4c78f496303d4 +size 12419401 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0edf40e3142c164212eea28fb4f2ebb15f40bb92 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs2 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs2/seed_13", + "wandb_unique_id": "deadly-corridor-fs2-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a2dc325b6659fe642223f7e0399cfb09b29567ce --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:26ea599e0a93daa045f5ff9f28be78df0f8b161ccdd1ad04b3ca7f499d97d935 +size 12419401 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ca6abcde6605e8326c2a0b04de9df52194843bc0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs2:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs2 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 2 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs2", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 2, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs2_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs2/seed_14", + "wandb_unique_id": "deadly-corridor-fs2-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth new file mode 100644 index 0000000000000000000000000000000000000000..02e18b317c8d8d4efc58f5002f5b6f3644c1ad99 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048844_25008128.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a98b6e49a8f87e63126461d3f1dc98e7b98db21282f9fd09df3eab13f96bafa1 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e76f5ce5e0ab6830646bd4291e92458f2b282261 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs4 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs4/seed_10", + "wandb_unique_id": "deadly-corridor-fs4-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..09e27e17cb8f4c4f100dcbd76eef55d81ee928d3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b2fb68b4ec2492ba1beb17d594f84b64f6ddad42bafbcb9ddc1fc65414257218 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..af760c2b18fdc8753f99563bfffe8e5f98b8e39f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs4 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs4/seed_11", + "wandb_unique_id": "deadly-corridor-fs4-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..28550b1f35c3e139647983d2c824ab33c1e5343b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:31ee7b905c21f99d4de1334b6f82ea07c82b048f52c2945fa7b646b0ef64d0eb +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dcabce1b44eabef6d572c644159185c34a6f8ec8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs4 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs4/seed_12", + "wandb_unique_id": "deadly-corridor-fs4-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..55072aa10f71bf0da76e05f4d01a4fb15c4b90b1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db67d1af02be7433904c17f56c78a61ff74174762d5eb6a9d667985e1da18138 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..62c4f725606164039b537c164deabdadc1ec985b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs4 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs4/seed_13", + "wandb_unique_id": "deadly-corridor-fs4-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..c943c1fba4c9d02fd58da759084171e077d7b117 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f6d48e1bbe9c9ee10020be0573ef56875596ac18d1cf1a90381c20703656e285 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..05b66f42027bc438800f681675e3ba85cc8ec7b8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs4 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs4", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs4_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs4/seed_14", + "wandb_unique_id": "deadly-corridor-fs4-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..097278c755b7be7ae05257dc5d0e2c496ad81235 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73788ea7aacdf1bab3c9c1a908988aaea77c17c7f31b2d97a93eda023f56e89b +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d42ece5b376911944ff2ce30675c4c00d8b9f0fb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs6 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs6/seed_10", + "wandb_unique_id": "deadly-corridor-fs6-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..44fed3c3cf7f35019610e45904f5e365528ab89e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c3956fd2d16c90e02d1b958d8964eed25a89f3f3c769155d1e0a8dfb65654597 +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..edc7ad31c484ae1ada530d3c8a936b6bfc275fd0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs6 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs6/seed_11", + "wandb_unique_id": "deadly-corridor-fs6-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..d899af112fa37331e772b6c38459529a51c2c8ae --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:45e69409cf5d0ce0929ec24add5e72d094b28d5c3fe923e9ceb3d93aa73d480f +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fc332c6ddb955e3f1a618d94e414d7f60b255cb6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs6 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs6/seed_12", + "wandb_unique_id": "deadly-corridor-fs6-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..6055724bc3b4d9748a90a0b954d945637e79e537 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:99267ada58e9d84ef24993501c3eb2da10d90c9a14b9c19f1028b230a3a4777b +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4f6ce59423df3c5a4353f1d286ad77a5f7e9c285 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs6 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs6/seed_13", + "wandb_unique_id": "deadly-corridor-fs6-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..355f8039f4612c7221446e83106dfb5ec88aaf1a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:87ce0c8189549d6473a3ecbef22bfd1eeea42654da6c7b4d5c0fa58885931415 +size 14287177 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0c0051697435d8bda9d3aa261d6711bafcd705e3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs6:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs6 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 6 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs6", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 6, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs6_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs6/seed_14", + "wandb_unique_id": "deadly-corridor-fs6-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..ddd3d2667de3f7fefabaf9ce69143491355ff0bc --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de956efe75f49daa91919f334a768430522942178e4ee02b92d250d18b2bcc13 +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..562b98a739ab41cd175691d730a987fa7b527577 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs8 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "284fe8ace24f0e8a40c03c5b559969abd7caeb29", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs8/seed_10", + "wandb_unique_id": "deadly-corridor-fs8-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..544c9e702a2b03258060f652c46d9c6947c651cc --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8593978461592fb6b14406f12165dce1d1e80f71f20746baec93e1f073b7ee5a +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cda2740f4f5ba93eab5faba115bfbc5ca8073415 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs8 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs8/seed_11", + "wandb_unique_id": "deadly-corridor-fs8-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..a084f4914efaf0796b130e287a2e5a6625d46254 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e4334505b037c82cf2e3b88bfa475a31fab4f1e2baa8c3edb4deedb6f7bd913 +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f23d01c9b1e2116987dd6fe5dfff201bd2ec89f1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs8 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs8/seed_12", + "wandb_unique_id": "deadly-corridor-fs8-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..bbea946f05d679048c3b62b2613e5c4e179315f0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16fee613664fd9b0a92a12b3918243b77019d2496d3d772ff5d031c35321bdc5 +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d8aa80aaad7fa0536fc7ada67109c954495141d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs8 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs8/seed_13", + "wandb_unique_id": "deadly-corridor-fs8-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..023fa9e455a96ee7142b8ddaff94d8a416e847b4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1f2e4a1075022ec2ef05d7c3f10fbfa5d1ac9d2f6a8428c5693a810ad7bb902f +size 15221065 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b207cf45458d2d7ae8693658a089d8e10628c51e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:frame_stack:uniform_u2_6:fs8:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/context_window --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-fs8-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor frame_stack uniform uniform_u2_6 fs8 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 8 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-fs8-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "frame_stack", + "uniform", + "uniform_u2_6", + "fs8", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 8, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/context_window/deadly_corridor_frame_stack_uniform_u2_6_fs8_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/context_window/train/frame_stack/uniform_u2_6/fs8/seed_14", + "wandb_unique_id": "deadly-corridor-fs8-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..1ff52401d7afdce339f4bd7fdc727641a25660a7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ef0433e44ead53ec77b3295ed94f71ea2286e9dd27ba6301ae8a9eae3c59a3ad +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6ffd373f9ced9c8e720145787bfc217566109a34 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride3_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride3 seed10 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride3/seed_10", + "wandb_unique_id": "deadly-corridor-stride3-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8150707db74576288cc56e971923ff2db4651d74 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6a96e2c141f39e926a23b07d8a20929e6dd96c131e9531d394db7a752d8116f2 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..919827df4e07f30b5f51e4732f0688b42987a987 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride3_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride3 seed11 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride3/seed_11", + "wandb_unique_id": "deadly-corridor-stride3-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..f561cd03a66d4bdaf6bc5233e61c117d529f57e7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6f85b5892d963c9a9ee6403b0b618bea8c9310acef038757ebafae637574933b +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d272037cab6d8a8932b1ca10b8681401b854c9f5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride3_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride3 seed12 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride3/seed_12", + "wandb_unique_id": "deadly-corridor-stride3-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..7429f368dd9b3574f65e22b7d2a0ca1fc37eff59 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:241eb557e75ee03e84633b5de18307d0f07001bc2390e6f214c128d7ff480d9a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..520e4aeadf8df56e87fe4851fcaf7fae827e845a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride3_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride3 seed13 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride3/seed_13", + "wandb_unique_id": "deadly-corridor-stride3-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..60e064aede64d4b9fef42b6fb61d2837a5377677 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7b96c176ac91babc1303d07c2df48cfc34b070ea2e6928617ae54df292bb5f31 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fa77148f2a77f8f767801c710eee330651585a85 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs11p6667:stride3:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride3_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride3 seed14 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride3_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride3", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride3_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride3/seed_14", + "wandb_unique_id": "deadly-corridor-stride3-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..df3993b47cf8c0180f85c4d4c3060ea6b8710e0c --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:98cca6c4b254f013cf8479076a487599fa9c96151dc47bfb3b9373d1c4c4fc54 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6bb9ec2ca55f3fc2982a222dc249d8bd13a65e37 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride2_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride2 seed10 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride2/seed_10", + "wandb_unique_id": "deadly-corridor-stride2-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..ca08698de5cae001c26ba3829871025a1e5472c3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d59492908050b287521496d255f6046e125a00166a20ba15192518a5f4a1e9f7 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..92310253cb68700a32c2c25a605b353086160539 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride2_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride2 seed11 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride2/seed_11", + "wandb_unique_id": "deadly-corridor-stride2-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..92dbe57da7d758af7c90a1ef27dc8de8a19fac46 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:34827356b458ab31e5e9f1124c3254421f1a7b31b8a5cd289b22df96c8030950 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2970202ed64aa7b7df8c73734d90c7ac057406c3 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride2_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride2 seed12 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride2/seed_12", + "wandb_unique_id": "deadly-corridor-stride2-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..984aad5b621575d80137f441141de8d2409c8a02 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0789debffbc2158dea25e556bdb73772c5eb10f77da5014260ccf2fdda267998 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..95eb5eb6d7fb834a6ddcce62a76515457df476a4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride2_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride2 seed13 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride2/seed_13", + "wandb_unique_id": "deadly-corridor-stride2-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..ea78e932c12a3a635530449587b2dc2a3070d476 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1ed063c7297d2183795c6d8ad622cba6b6b9ae6271ad4d42eb2f77572f6f7021 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7fce6883484f9f8208a6dcb02c4fb822d546f53 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs17p5:stride2:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride2_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride2 seed14 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride2", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride2_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride2/seed_14", + "wandb_unique_id": "deadly-corridor-stride2-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..fe0ae5d0caa13847de53f094b6642827949f05e6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5fb3f645da0da15db5eaebc95689647411b9f47e9d14fa5e3cc74582013666aa +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a26dbb2a1fa4b7098641dd1f673ccd33e80893c --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride1_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride1 seed10 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride1/seed_10", + "wandb_unique_id": "deadly-corridor-stride1-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..30338f72643faafee4f6f1cc91cbb6f6640d0a11 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7309dd409e97f535ad3d543adcdf3ffae8c2c83f0e43af600cdeb471bff28c37 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..effe69e12de57840fec770b427573756215840e4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride1_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride1 seed11 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride1/seed_11", + "wandb_unique_id": "deadly-corridor-stride1-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8c77a9a3e6ea133db4e62a887addbc34e45da387 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0cfc2973b2f9fe0f7f46b2900655c4c1cd9101de06dd45e6930ebd4aa6c0e418 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b36418bf47aa4042a959030a4f8cb5103755b04e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride1_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride1 seed12 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride1/seed_12", + "wandb_unique_id": "deadly-corridor-stride1-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..5d57c9408310b755d6dab0e22aac7cb22eee0df8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9edca8f0c1f79561361422c9857a8714af99b145a21328aaf08616e643086a7e +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b25fea914287c10ce7b798f62a96f8084699d8ae --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride1_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride1 seed13 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride1/seed_13", + "wandb_unique_id": "deadly-corridor-stride1-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..1513c8bae0263f2eae155689a45c9d6721c64a00 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ae5414a397993850e8de6c7e5dd26e3bb01f1fa2d7a39a999703eb5c8f7ea159 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..30447e6787a6278be6ba2d1a05d1e6287ce270c0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs35:stride1:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride1_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride1 seed14 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride1", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride1_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride1/seed_14", + "wandb_unique_id": "deadly-corridor-stride1-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..2cd6ef4e2710e7119f64c32df0166a836f0b0553 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:516353dd230dccfd348b25decf4d65668ff1aaaf0cb66b6259caba748ffb8d3b +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1cafbd018b5999cce9e82931083b0d411d3efcd0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride6_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride6 seed10 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride6/seed_10", + "wandb_unique_id": "deadly-corridor-stride6-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..157dc8bbb49fa70ac622b2288d23c926f301dcbb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cef12b50d701560e9f9fdc86da74bd028d1c02a49f4cbfa19e49e636edd81efd +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aea2af27e7f9b8ea4017f1e8ad91b82b8817b6f4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride6_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride6 seed11 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride6/seed_11", + "wandb_unique_id": "deadly-corridor-stride6-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a9aedde1de016792b07ba8724b154923779b1319 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6e990229272411cf177d56e5c97cfe85341497828a3848cace8c6055498b1a72 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..08671700727bf547c19c32c984bbbb91ecbd7d27 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba7e0ff1b998eeac1e6704b39992ccc61eb2e116e00f5775f601f89e693090ba +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..693849000c21500266b57655f47c6cb035c12024 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride6_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride6 seed12 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride6/seed_12", + "wandb_unique_id": "deadly-corridor-stride6-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..268747f34fa06b35cd24b890db56167fcaa686c4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:60a91a2f878f96d430206a61288c72faed3843303fe74fb0b3bd6872b0482b7c +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3693aedea388197ffc0c4b7f6a365b28457022fd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride6_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride6 seed13 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride6/seed_13", + "wandb_unique_id": "deadly-corridor-stride6-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..153fb5165d5a9504c64bd30b69a463ab8850fb82 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4b52dea74b39dd90665cf3c718fffea3fecc8fe011807c726ec60f46e41dc031 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..94520d116deec895ae6db0f45ec1f6001d77ce83 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs5p83333:stride6:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride6 seed14 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride6", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride6_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride6/seed_14", + "wandb_unique_id": "deadly-corridor-stride6-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..90c166714690712ce8fecbcbbe4e0d1d55f8c1f1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:bd487ed550427d00957d80586680730018d886bc3ac682d492710f9c6dc5fc0f +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..18f54bd2b91e2cf4092237216944b64a4b422324 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride5_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride5 seed10 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride5/seed_10", + "wandb_unique_id": "deadly-corridor-stride5-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..9eb567edb2f2a627bcb22d42278cc8ed8acd2467 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:716de7baf94e5f975dd9d9c5c88840afb77269607c16bb718b5afa7db146c98e +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1474c6382cd87f5530682ae59e1fb7f4f26925ca --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride5_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride5 seed11 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride5/seed_11", + "wandb_unique_id": "deadly-corridor-stride5-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..8ea52cd659aaccacd48dce504eeb71a20e2ee321 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:318304e07d049d916671413cc0f4d83581b8fcc99535392b3ebc63d64c79fda1 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..49a8c77970c6dfdf0f9f8924594a3760b45cbc91 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b642d6aa7700d7879b7f71898605445ef82ca5e64a99e65c9311ddcf2658d10 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2f04007d9a98e2336b524274b871df1357c3992b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride5_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride5 seed12 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride5/seed_12", + "wandb_unique_id": "deadly-corridor-stride5-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..398be9744dac29ba93a091c1edfdb918fc594555 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:290a9f68952e157f8648d8cf57998a0190ca46942888ca04bbf81124638da320 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f84dd178d84a61e51903d5ede8f5fb8ac4d013cd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride5_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride5 seed13 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride5/seed_13", + "wandb_unique_id": "deadly-corridor-stride5-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..e39abf8ca15d711d099a651120ec2414bdb40ed4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8402b7529248a32e243b65db8e9139e4aa05dcdb964acf8ccb5af1970a77d906 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0bb13ef223af8f6e28166609a3133c5bc917fa6a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs7:stride5:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride5_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride5 seed14 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride5", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride5_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride5/seed_14", + "wandb_unique_id": "deadly-corridor-stride5-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..ec8ddfec8315ef2ee5f874043d1e35cb8cf3d2eb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:496fadec253700d2973d46c5c3708242e88e7e078b1662d72466991847afd38b +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6084f8fb6ab641dce19d99b98ac2e62c091c8e8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed10/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride4_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride4 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride4/seed_10", + "wandb_unique_id": "deadly-corridor-stride4-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..1db565f8744bfcc3736d86d4140a9def85080813 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4cb68750122d19a45b3c932f2ea86cf7bb247210170850910b344676b363f77 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..149ad85bef1ebca8c43841416f42b3f2f21e0fa5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed11/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride4_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride4 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride4/seed_11", + "wandb_unique_id": "deadly-corridor-stride4-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..c86e28a92edeb0266a10f39718964d7a5ff63ff8 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:95c8d29c739e5016e48be6d4dc09295a71293a85d09cd4986d61f7f59bc3437f +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..89b8e2e40665dccbfefa92fb2c902b40be98b1cd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed12/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride4_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride4 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride4/seed_12", + "wandb_unique_id": "deadly-corridor-stride4-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..82e432aab9e706d8d9eacad4efc5c44760e0b392 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:47f7a05e9f87d9906715b80fc2eff59b75d2b565b04e3247d2615b642cf7833a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a975889aa091477e4439727e94655b73953e308a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed13/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride4_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride4 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride4/seed_13", + "wandb_unique_id": "deadly-corridor-stride4-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..d1dfba9af025c1d082a85c6bfbab5208742fe762 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:60d0acf989fdb4e8dc037830f8ed9ffe2da821238479da23a6952b021e214f1c +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7ae9d0cb90c8ea1ecff2fb525ffd6ed00d7eabe7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:fixed_l4:fs4:obs8p75:stride4:seed14/config.json @@ -0,0 +1,261 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_fixed_l4_stride4_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 114.28571428571429 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride fixed fixed_l4 stride4 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_fixed_l4_stride4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "fixed", + "fixed_l4", + "stride4", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "fixed", + "fixed_latency_ms": 114.28571428571429, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_fixed_l4_stride4_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/fixed_l4/stride4/seed_14", + "wandb_unique_id": "deadly-corridor-stride4-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..98067d997c023629586d61176badbd9bfd3468e7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1953425b748111264a4d9f04638ebbfae8c2b74f4d11b0aa712a7b12a735546c +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7ef127731dfb5f316fdc29f339031ae62adce579 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride3 seed10 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride3/seed_10", + "wandb_unique_id": "deadly-corridor-stride3-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..7f48fa9d3c608b7a169ea58938ee4115f5a57470 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:973d2b152476dde80b0ab1be9d66ed11a6e44a37529e8a8a8aa7d0b5bd2877b3 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..7d2ddd96539bcf7fec5a51ed01dfdd3fbebf04c0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0dceaa631c841ccfac200ea75455d63083c83e4e7b5d97e3664e717e94e21c77 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8b1206c48405e3ac6ddfbd4b2b98194241cf3586 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride3 seed11 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride3/seed_11", + "wandb_unique_id": "deadly-corridor-stride3-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..af36c7e90ee7b88f2b628244506d887dd5206f0e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:97fbe934c9a28f0fe15547290bb6ce7663b31d5d3976db02c8247ca18853a9a4 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f1f679edfd6a01994941242cea9f310831075bec --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride3 seed12 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride3/seed_12", + "wandb_unique_id": "deadly-corridor-stride3-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..75e275c3afd0f828e72288fe726c7113bafa2085 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:380ee41d7dcb3407473049516e7e13514d3bddd981a164065767dc1ea31f660a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..ba02c44994170e6c77e747025219ab11bf796b9a --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c395758de974116c754acd6d36cc685be79c71115bd1e9f9d3a33891a7c4f9a4 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..581349c87491aed3f913c9d6e3e41071d9bc7b0b --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride3 seed13 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride3/seed_13", + "wandb_unique_id": "deadly-corridor-stride3-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..dc6dd78a4097eb553e9bca509ae74efb5c5fa702 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:345c4b62af029741aa4815321434e353537482184c2cc4540fd17e343b07413a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..53ce3d6d3b6afe5634baa441afdc3ace9e2b2f16 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:44c7a5d73aa55b872d6c9a8c7ebae7f425d6f9dfd8639ba9d6e3edafa7a99c90 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dbc472c00915b1d68afa023192bfb76fad55ab3d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs11p6667:stride3:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride3-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride3 seed14 --env-fps 35 --obs-fps 11.666666666666666 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride3-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride3", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 11.666666666666666, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride3_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride3/seed_14", + "wandb_unique_id": "deadly-corridor-stride3-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..93e3233ae251e3db294c2ef0d82b17f9d79809e7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c7c054e7d5a7f05fd6d4fe310eac21d1f7e4f137807af7eadc8cb7f2626425e5 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..0462165f8d6b42e4f8f6e9d76a8a1696b1150900 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9a32eda33529cb6388c8a0f22fb6e69a0bb04f9aa9c4a96887d55d0eab553989 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bf585c6b5cff229ace7aa7edbf4b761259083431 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride2 seed10 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride2/seed_10", + "wandb_unique_id": "deadly-corridor-stride2-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..98e6e996aa7ff23bc9c35e2de143824503e7d716 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ac301afc3f9f1d81404526df00ca67d3a08c9386ee1a635c3228c4241ca3809b +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3ea7d1b3373370a4a92fcabf22a31584c37cb4d4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride2 seed11 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride2/seed_11", + "wandb_unique_id": "deadly-corridor-stride2-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..b08e37d7803f52b0225cadf384b91d4c7294fabb --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ba4c894c3b78221c5ca8749b99b12956819b6334d3aa2294255b85411610e1c2 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a5ec03670bb2a64368a9aeefeb39bd6af500852 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride2 seed12 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride2/seed_12", + "wandb_unique_id": "deadly-corridor-stride2-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..32c82b23e8441c7b41239664aa0dea4c60a406b7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:75b5126416f4ef7836e4e6522861f3d0f8247beb6a390655dc3cc0b2bbb00181 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..81636aebd143a32f9a3b18b8a4a2d9e191a320da --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride2 seed13 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride2/seed_13", + "wandb_unique_id": "deadly-corridor-stride2-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..bfb2c13f33975cd1a3be5a77eb25cdbcfa7e3bf5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c1c3e640f12754d65522fe225bd2c2bf5560f8e6107967765daeb5ab96596b24 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e2c29ca0adc98cff7894133c105ef5e43283936e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs17p5:stride2:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride2-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride2 seed14 --env-fps 35 --obs-fps 17.5 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride2-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride2", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 17.5, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride2_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride2/seed_14", + "wandb_unique_id": "deadly-corridor-stride2-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..058e696ff478c78fd227df24a2b7d399f2349821 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:635e5c25302218b33a110749c1cdb9bbc4e8ff102a8cc5ebd20c8cf1d01caeeb +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0cb885924b7a17ae261a52d00b76dd3855f4b519 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride1 seed10 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride1/seed_10", + "wandb_unique_id": "deadly-corridor-stride1-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..0d1881861eec64333786a17ec4950c08b7aaee47 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:985a0b0f23a5c3407c80ded565e09be6307c2026dc926f295aa8502ce40577b1 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e27ff45819ac88e3604d46e6ce4d2761b677fc24 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride1 seed11 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride1/seed_11", + "wandb_unique_id": "deadly-corridor-stride1-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..a7bece78a54132527a9bd44c621b56a274ad21c2 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:13f61257d438dd23e2893855b37c2ee3073e197f81818100add49712b86eeee3 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1c4667f77b7ce0bbc592da835700dba32354ab43 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride1 seed12 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride1/seed_12", + "wandb_unique_id": "deadly-corridor-stride1-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..48d563b2957fec5ddb9b5fc87520db22aa3282e0 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2680f2f3a0ecdbdb334341066e4dd8a1cf353f5bc2f3226a1efc9086003fce21 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2b3d5999b5609118017e0a6739bfc3a01ad495e4 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride1 seed13 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride1/seed_13", + "wandb_unique_id": "deadly-corridor-stride1-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..6464421255d058ead616b17e6ac13106bcca5ef5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:671d478d7196d3eefedcf8c037a39ee2f9209c06912e14f0db9763625cbc405f +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea9a38d3c64c51e8955366972b4aedf6f8ecbaf7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs35:stride1:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride1-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride1 seed14 --env-fps 35 --obs-fps 35.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride1-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride1", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 35.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride1_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride1/seed_14", + "wandb_unique_id": "deadly-corridor-stride1-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..43448322b3994587303d7dcb40865c3522afb840 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:35c2338655b964ba769b534e53511c06b163d35b70789b41d3d97ceb40f68bf0 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..bc5407d03fbf3bedd53e1c0b4f12ab67664acd86 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed10 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_10", + "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..2190ccdbce795b45b542ebc764f939440cc11fcc --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9e6f869fc1ad5dbd45288fb582ba178f075eea188ad309ffb00ca5fc1db357ff +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..253add9f647b7974b3f346061bce5238cab78186 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed11 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_11", + "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..764f5708fa5d9ffaddb9d1b712f7f0d1e2a80661 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d9ee24783c3136de50dcf7e259e05e51331de8692287f647acccb26d710ba651 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..23b02672580ce91fce427468bbb663d8fa21d6d2 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:16b5d60c8a69f328b424b38bba9f89c28e6a20ad1715bfce49dd9ecfc413652c +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..46f9c81e08d32052d11c1abdc995c07c9e43092e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed12 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_12", + "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..9002d1e5c548ee17412ff2e2b55daf1048a29240 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0bc5c89dd97295a9c816c93db075e96b3ddd261bf340b3ada6369f366e472001 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..3d7c5aa00d2d22b9c5d165a561a33fc2f777a3e6 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:e5be2681fe4cd02fbd68f4f6523eeb74458f6de351942ce4dce2e19f73b1af3a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c40338cd4bf795b94bdbc17c9fd068cc7cf8e7b1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed13 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_13", + "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..f6ebf9d2855db8725026092fd451574e307d5455 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:10df9186c041c504a8ac9cafa2fabe44c444cc964a96a4f197d414491c934482 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..cbc3852056bac6960a54cdbbbbb2b2df296c8d0d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs5p83333:stride6:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride6-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride6 seed14 --env-fps 35 --obs-fps 5.833333333333333 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride6-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride6", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 5.833333333333333, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride6_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride6/seed_14", + "wandb_unique_id": "deadly-corridor-stride6-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..20d22fcb1e3f704adc6057ec7dcc1f44fda79b25 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:08368a81ee4e1116127fcc95a7263ca9b375cb624b4d80cceacdb3065cda65a6 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1995bb37c11516bcc7a3c514641a94d962a844a7 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride5 seed10 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride5/seed_10", + "wandb_unique_id": "deadly-corridor-stride5-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..a4a4267a9b629b0333ab47e1bcb3c8d7b03cb7cf --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:25ee818bc8067847f89f397b6fb7d44f038a4e064c1267aea9abbaf82ca8257f +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8fe6cc8edf7c534cf1e6e979333efb4dc078b701 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride5 seed11 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride5/seed_11", + "wandb_unique_id": "deadly-corridor-stride5-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..2aaae98bf57b2c2eb82eb1da99826fa6d2d1732c --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:92d9b590aaefdc050da18a23de54226f467bbc85859e49a1f967306811d36c6a +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4e10b555045dc27af73c78b3f4736c1bc6b5c00d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride5 seed12 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride5/seed_12", + "wandb_unique_id": "deadly-corridor-stride5-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..f4d15e45b7938d34eef3e68f979082e557938962 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:73be58dc658781d947739a61fea15385bcd078f71510c1d1302db69589f70cf1 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth new file mode 100644 index 0000000000000000000000000000000000000000..cab6b32a7a3c1a828d738bdbc9ec22ab97c116cd --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/checkpoint_p0/checkpoint_000048840_25006080.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f57f6c6c3ec4952eec8b10771786e094845b8eae614ede7fbddd7be85622a0bf +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f6e912956131d639ad98f9bd12fb55b1a8bc4838 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride5 seed13 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride5/seed_13", + "wandb_unique_id": "deadly-corridor-stride5-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..6b5779d8160c189625b708138914a7164ce8b2c5 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b902ebf4027624277f8c24fa7191d445d42ecb2bad9f0a3a2d4d0474458fb99 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..24801c0346f9d43585863dcc48cc9576a31964f1 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs7:stride5:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride5-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride5 seed14 --env-fps 35 --obs-fps 7.0 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride5-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride5", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 7.0, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride5_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride5/seed_14", + "wandb_unique_id": "deadly-corridor-stride5-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..5f564ad6f7a314020eaadc30bb31d558ad2b1621 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:526a7c684675f9fd842d4843a6622a28ba8f5439fc5f60c371e77f16bb677b79 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..d0ed6a747b7b01cb0a0e8ba98eb640e9490386ed --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed10/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 10 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride4 seed10 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed10" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 10, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed10/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride4/seed_10", + "wandb_unique_id": "deadly-corridor-stride4-uniform_u2_6-s10" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..0650b50e7b357f0d8f0fc21fe266057c4984a81e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c98faf087c0c04c54b97043b3c014ba2891c7e67cceff26f5c86f7bd5d0cacc0 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8b18104efcb55bf82d34abf4b0369d28076fed5e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed11/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 11 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride4 seed11 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed11" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 11, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed11/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride4/seed_11", + "wandb_unique_id": "deadly-corridor-stride4-uniform_u2_6-s11" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..57ac91aa6cfa8b46531886b981d754996bae8e3e --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6379a09680e40cb341cfaea7920fb8668b4422413263c6a97553a7ded3a541ea +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..7a1549d61f37821c749358e6cecd9d3e0730aa9d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed12/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 12 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride4 seed12 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed12" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 12, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed12/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride4/seed_12", + "wandb_unique_id": "deadly-corridor-stride4-uniform_u2_6-s12" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..c81be7361aaa964145f26082c1e15c49c590889d --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f4977bc8fed1902fb662ba69b9aec7520e7588b4a4af9c6addae6c50d4c23919 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0a9f2b3f4444da06b462e7649d0d8499f734dc6c --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed13/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 13 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride4 seed13 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed13" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 13, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed13/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride4/seed_13", + "wandb_unique_id": "deadly-corridor-stride4-uniform_u2_6-s13" +} \ No newline at end of file diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth new file mode 100644 index 0000000000000000000000000000000000000000..2c7edfb2e0bf5ca44d711e34a930442973e68f8f --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/checkpoint_p0/checkpoint_000048836_25004032.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2c57749b5c5466c524a139d2c05688427d3a73af2a16751dd573b4453159f5a2 +size 13353289 diff --git a/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a4722c51a31e58f09731fd6a07eed7c57aca12d9 --- /dev/null +++ b/latency-aware/deadly-corridor/small-policy/factor_sweep:deadly_corridor:observation_stride:uniform_u2_6:fs4:obs8p75:stride4:seed14/config.json @@ -0,0 +1,263 @@ +{ + "help": false, + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": false, + "num_batches_to_accumulate": 2, + "worker_num_splits": 2, + "policy_workers_per_policy": 1, + "max_policy_lag": 1000, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-06, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "lr_schedule": "constant", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 600, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_simple", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": false, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "fps": 35, + "eval_env_frameskip": 1, + "no_render": false, + "save_video": false, + "video_frames": 1000000000.0, + "video_name": null, + "max_num_frames": 1000000000.0, + "max_num_episodes": 1000000000.0, + "push_to_hub": false, + "hf_repository": null, + "policy_index": 0, + "eval_deterministic": true, + "train_script": null, + "enjoy_script": null, + "num_agents": -1, + "num_humans": 0, + "num_bots": -1, + "start_bot_difficulty": null, + "timelimit": null, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "record_to": null, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "fixed_latency_ms": null, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_latency_raw_frame_values": "0,1,2,3,4,5,10,15,40", + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "max_episode_steps": 0, + "simulator": "cpu", + "envpool_batch_size": 1, + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14/episode_metrics.jsonl", + "vizdoom_init_max_parallel": 10, + "vizdoom_init_lock_timeout_sec": 5.0, + "vizdoom_init_lock_max_wait_sec": 60.0, + "vizdoom_init_retry_jitter_sec": 0.25, + "command_line": "--mode train --algo APPO --env doom_deadly_corridor --experiment deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14 --train_dir results/checkpoints_factor_sweeps/deadly_corridor/observation_stride --restart_behavior resume --device gpu --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 32 --num_envs_per_worker 4 --num_policies 1 --batch_size 1024 --rollout 128 --recurrence 128 --num_epochs 2 --num_batches_per_epoch 2 --worker_num_splits 2 --learning_rate 0.0001 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --exploration_loss symmetric_kl --exploration_loss_coeff 0.001 --value_loss_coeff 0.5 --max_grad_norm 4.0 --rnn_type gru --rnn_size 512 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --async_rl True --use_rnn True --normalize_returns True --normalize_input True --latency-type uniform_distribution --min-latency-ms 57.142857142857146 --max-latency-ms 171.42857142857144 --latency-seed 14 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 32 --eval-max-steps 30000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group deadly-corridor-stride4-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep deadly-corridor observation_stride uniform uniform_u2_6 stride4 seed14 --env-fps 35 --obs-fps 8.75 --res_w 128 --res_h 72 --wide_aspect_ratio False --frame_stack 4 --simulator cpu --disable-reward-scaling True --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "doom_deadly_corridor", + "experiment": "deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "worker_num_splits": 2, + "num_workers": 32, + "num_envs_per_worker": 4, + "batch_size": 1024, + "num_batches_per_epoch": 2, + "num_epochs": 2, + "rollout": 128, + "recurrence": 128, + "gamma": 0.99, + "normalize_returns": true, + "exploration_loss_coeff": 0.001, + "value_loss_coeff": 0.5, + "exploration_loss": "symmetric_kl", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 4.0, + "learning_rate": 0.0001, + "normalize_input": true, + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": true, + "rnn_size": 512, + "rnn_type": "gru", + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "deadly-corridor-stride4-uniform_u2_6", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "deadly-corridor", + "observation_stride", + "uniform", + "uniform_u2_6", + "stride4", + "seed14" + ], + "eval_deterministic": true, + "res_w": 128, + "res_h": 72, + "wide_aspect_ratio": false, + "mode": "train", + "env_fps": 35.0, + "obs_fps": 8.75, + "latency_type": "uniform_distribution", + "min_latency_ms": 57.142857142857146, + "max_latency_ms": 171.42857142857144, + "latency_seed": 14, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 32, + "eval_max_steps": 30000, + "eval_raw_reward": true, + "disable_reward_scaling": true, + "frame_stack": 4, + "simulator": "cpu", + "episode_metrics_path": "results/checkpoints_factor_sweeps/deadly_corridor/observation_stride/deadly_corridor_observation_stride_uniform_u2_6_stride4_seed14/episode_metrics.jsonl", + "env_frameskip": 1, + "eval_env_frameskip": 1 + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "output_dir": "outputs/factor_sweeps/deadly_corridor/observation_stride/train/observation_stride/uniform_u2_6/stride4/seed_14", + "wandb_unique_id": "deadly-corridor-stride4-uniform_u2_6-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..631669629a51c2fc682d4ddf2a5fe09a2b49aa70 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fa62c69ab6a7f04ca1b35bfc3b3b54cbf7478cd756a5a1ecc506ae880ac1ea09 +size 7937341 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..53868aafdbcf2142ebddd6d9b1deec4ec9e56f2a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs10_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs10 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 10 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed10/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs10/seed_10", + "wandb_unique_id": "demon-attack-fs10-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..58b66f3ec274766d0351c85cc61ff28d446326cf --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0e34f56ada7e215df1dc446c4cec6a906174d066e7b37dc412bd5cddfdaffa62 +size 7937341 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..da3012e957c6526cb9dc50cf20a11341c84f7ca5 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs10_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs10 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 10 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed11/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs10/seed_11", + "wandb_unique_id": "demon-attack-fs10-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..dcc710a07d48c5470380397bbee35c32a875218c --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1021c6fb442f5db6072252dc1c429ca789838790d9b4587e355af79307771b0b +size 7937341 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2a336a68146c52a6e7d255741147ecf54383a56c --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs10_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs10 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 10 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed12/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs10/seed_12", + "wandb_unique_id": "demon-attack-fs10-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..c5dd2843796b356f1a6e4d4ec163021ae8c4cc21 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b400808ba0f9cfe64e1e78ba94169fc3933afbb94d6610727497f3b61361abbf +size 7937341 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..3e82fe9ea2435ce9b55d5432d3fad6129d26c487 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs10_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs10 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 10 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed13/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs10/seed_13", + "wandb_unique_id": "demon-attack-fs10-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..3ec475b6dc97f91e7a3163eb3666a800abf680b0 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:4941e1ae1251a99765bbc06fad1cf5e90f50640a7163fd59dfe47f6d8346ea1a +size 7937341 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..0f6699c92424ecf8196747a47d00817091ddc719 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs10:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs10_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs10-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs10 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 10 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs10_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs10-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs10", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 10, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs10_seed14/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs10/seed_14", + "wandb_unique_id": "demon-attack-fs10-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..f4288fff567da1e97ac38fbaecdbd6c16b39ae50 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a5f9fa85752b6c01f32b042b31913ed6df5b8b986b8a93315f4c5bc2e4e89865 +size 8058429 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e17b2dd679df3f062ca42e3a5f8b22c2c23daf8e --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs11_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs11-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs11 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 11 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed10/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs11/seed_10", + "wandb_unique_id": "demon-attack-fs11-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..26e39aea03f7ffee487fff83a0e4baa52345830f --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d4620ac7005c361c0061ee5b6c345e528e74e045b0a91d98637bc10c11ea0298 +size 8058429 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a89862b7134c54fcbf603e64beb01695ec2ede37 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs11_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs11-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs11 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 11 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed11/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs11/seed_11", + "wandb_unique_id": "demon-attack-fs11-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..ea4813bfe0aa16cc0c8954ba0b5e0ae56830fb93 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7266f49c5086f33c903f2abdd3a963885070246ee79011304b7c349733b0793d +size 8058429 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..62ed4a7acc6c3901512eba35e80751d3e993914a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs11_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs11-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs11 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 11 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed12/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs11/seed_12", + "wandb_unique_id": "demon-attack-fs11-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..4dee8166a6a8dedb6792b5f2c0e28eed850c92f7 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:618ebce8c2a78c89e2483bef5362a0d11c976b8d9b06e5e414055f8f56878710 +size 8058429 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..39d7b688d19b70f2aeb2b30f6c85393a4316688f --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs11_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs11-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs11 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 11 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed13/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs11/seed_13", + "wandb_unique_id": "demon-attack-fs11-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..a5d3b9f38a75eca79405a0221ec32ab191b1a696 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:742e3dbf0187ebf6f5ec0ed967936ed34b36e38a5ae2f4756259a6a636e1ab1c +size 8058429 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..aa5098574dc7d68b2872f6e1c0bcbb7f5afdcbd2 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs11:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs11_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs11-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs11 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 11 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs11_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs11-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs11", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 11, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs11_seed14/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs11/seed_14", + "wandb_unique_id": "demon-attack-fs11-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..1dc84cbc4a455e4db90b824b17b21883453e7e76 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ff3c15923b3d522f6c96bc5749566ccfbef0db40f5ea1aac1ad1dd800312ec99 +size 8179517 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f4c4c762f2eb844bb72fd85ea7bd66bc4cc6f35a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs12_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs12 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 12 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed10/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs12/seed_10", + "wandb_unique_id": "demon-attack-fs12-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..1abdabde50464d0ce0fbc0259cbc7b99c0d2ef2f --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0f0142c4314b136fed0545a87d0a611c936b3c5e5fc986381209b280c0a1954d +size 8179517 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..849b94558b5773a8caa2bf92a925df489c401321 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs12_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs12 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 12 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed11/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs12/seed_11", + "wandb_unique_id": "demon-attack-fs12-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..ed6a6056b5935f2dd7fe9dd86132023bd3d8c6d2 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:293641d3fcebac8cb226590ecc1116e9c6e464451dabc50a9990c90d277e768e +size 8179517 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..851a717aa817e5b3f317c0002ff2cc989ff31d52 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs12_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs12 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 12 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed12/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs12/seed_12", + "wandb_unique_id": "demon-attack-fs12-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..9c94e12a3dd641c3e5e7403ad2f2c14a00f3777a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1b2b675f1e6503110285a0596275cb4db86381aa47270694a9ac6554252cb832 +size 8179517 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4d27b2396caaab6e5950be4e4eefa69db7363c47 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs12_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs12 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 12 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed13/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs12/seed_13", + "wandb_unique_id": "demon-attack-fs12-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..c1cdc53ee51ced2cb7302b1a4fbbf374a427bdb5 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:22aa6b7bfe28a36085a21b90c62a0ee406c68fdf04eeea29288ddd5fc0bc9f3e +size 8179517 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..dbb02f2d9b527cec54fa8962f71e37ca4ee18525 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs12:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs12_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs12-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs12 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 12 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs12_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs12-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs12", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 12, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs12_seed14/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs12/seed_14", + "wandb_unique_id": "demon-attack-fs12-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..d755a899a7f4aa91b93331ef31eedfbf042bbadf --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ea4958b4207a4e3cc5037097dbeed4fda84d3e88c26475f9834612912821f38 +size 7210813 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..856f54bb5ba290d442e325ba1f05038dddcf6caa --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs4_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs4 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 4 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed10/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs4/seed_10", + "wandb_unique_id": "demon-attack-fs4-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..3320a3250e224b5d95cdffe9b42ba2bada25f770 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c44447f44fc7ddd9febddceecc18f2766ee93cdb52079a5a5914fe96d4ed6c70 +size 7210813 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2182d80718b1389edd4ebe96d9fadc040a5fc25d --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs4_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs4 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 4 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed11/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs4/seed_11", + "wandb_unique_id": "demon-attack-fs4-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..01e8636d90ee91bf891c70d330833eefad054702 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:073865776915e91c84ba7c1e09680170126d3c1ef0032a80ab88e051309c7b38 +size 7210813 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..ea2d27869b560eb1240fd0ca270248f48349983b --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs4_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs4 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 4 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed12/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs4/seed_12", + "wandb_unique_id": "demon-attack-fs4-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..29957fb43d0e58c30df80477a9b3ddddaed038fb --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:af07672812769bb175a89af787703bd1a1e04db5693d3cc8ce2502001e57d5c5 +size 7210813 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..893484c6dad35432ca107ce50ff17875d9ffc9f3 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs4_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs4 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 4 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed13/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs4/seed_13", + "wandb_unique_id": "demon-attack-fs4-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..96d58271d135cc30d8b81d75201307c18b0b9558 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b43bccfb621f7a3dd0004652490ad3093ddb4b3bc012b1ae606419a438c2e263 +size 7210813 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..1a069dd3b145fc326aca8da6570f81dce9726935 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs4:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs4_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs4-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs4 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 4 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs4_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs4-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs4", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 4, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs4_seed14/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs4/seed_14", + "wandb_unique_id": "demon-attack-fs4-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..46687018231e359215ab03b751498a2e64e101f1 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cd80ea69cff906b6f417c1c873383c3c98cc5fd48de42f39a98db301fb639eb3 +size 7331901 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..fc278107e33cc5af4d262082b9a08da869ea1c83 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs5_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs5 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 5 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed10/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs5/seed_10", + "wandb_unique_id": "demon-attack-fs5-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..a903afda72584d600810dec9fd2d1eedeab6517e --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:be8fd51292052467e8d62e4597d8ed34257698c794c1575977cb8e4c526ab452 +size 7331901 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2453631f0d254aa6b57f6de866de84b03f7401c1 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs5_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs5 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 5 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed11/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs5/seed_11", + "wandb_unique_id": "demon-attack-fs5-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..ef4f3774786940fed3fcff4f473bc1bf00e36634 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ffcaa4b9aba51125b635640a2d23b74f97240a1987ba797c474ae2d6cccc1e75 +size 7331901 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6adc954fc20cc96c1d23488e7c0a36491c231823 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs5_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs5 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 5 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed12/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs5/seed_12", + "wandb_unique_id": "demon-attack-fs5-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..4a6c5943dab9a2807a9997219a547d6431e5f2de --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1cc910d9d8fb574ea3b976bce651f50aef874807a050bf3e34c69f1c7433015c +size 7331901 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9bc0ec0885850096ab5a42131d68d6e02cd0618d --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs5_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs5 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 5 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed13/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs5/seed_13", + "wandb_unique_id": "demon-attack-fs5-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..ad8c944debcaa09a76192b7f14d161a6e292091b --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cf1013e8429eff8b385182a0fc51e5ddcea53bcac407ceec80e538ed6d36a6a4 +size 7331901 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..07524f12a61b63811c707e006ef87c926f2a47a8 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs5:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs5_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs5-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs5 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 5 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs5_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs5-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs5", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 5, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs5_seed14/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs5/seed_14", + "wandb_unique_id": "demon-attack-fs5-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..2cd3df9e8ad3b5c5d43cdd688f290d81b149d65b --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ca7c44356f0f05d7cbfc26029df80477dd56e79c13332d3b8411988463c67b07 +size 7452989 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e48ad79cb40c01df91b59b63a321e8efcf841925 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs6_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs6 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 6 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed10/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs6/seed_10", + "wandb_unique_id": "demon-attack-fs6-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..8eae64305ce3cd809cfa351ebd731f3c12fe96ae --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:50a633bd911b269e2a55f5d676f2d1b2f645603122b18aefff272ee6f9c2e79e +size 7452989 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b226cae0cb43ac0bd50b727e931ae4b48d78fddb --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs6_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs6 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 6 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed11/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs6/seed_11", + "wandb_unique_id": "demon-attack-fs6-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..c86e24d924361bbcd39de686dc6eb80a42fa2c27 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:caf3de51addf298894d1ff1ef7c9b432421f76871985618a640b64c7150aa6ad +size 7452989 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a857dc7125c2366a4b6ccbd87f72d8e976175f9 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs6_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs6 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 6 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed12/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs6/seed_12", + "wandb_unique_id": "demon-attack-fs6-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..306f99878358a496ccd1e5c3c640c4895a2d2132 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1e97be74a3156ad4c7eff111db5a9abee732fde0752a4b5ea51b07bd86aebe01 +size 7452989 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..357e579f5b2b7287ece018574474f75569ac35c3 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs6_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs6 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 6 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed13/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs6/seed_13", + "wandb_unique_id": "demon-attack-fs6-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..33bc0929a5ba8c4ecdad97db1fd114de7840f408 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:71e357663121280dbc30b519aecd3ef8e651c6387d96a98eb8c93735692192c5 +size 7452989 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..43a6207e45c9c0f198a1a4765a2f63a0d093afe5 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs6:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs6_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs6-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs6 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 6 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs6_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs6-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs6", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 6, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs6_seed14/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs6/seed_14", + "wandb_unique_id": "demon-attack-fs6-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..a36669fa4b074c1b5498ee9a1e428c992e1c1bc0 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8bc7a5bcc9ccd9cedfff86c72d324610b0c7c17281b05e949a2206834c35c9d2 +size 7574077 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..8b0a9b860e752d736e177a8b957a4524368532cc --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs7_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs7-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs7 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 7 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed10/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs7/seed_10", + "wandb_unique_id": "demon-attack-fs7-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..b641f9e20050ff18857c3dac79dbed68c98ff835 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b441cbc13df11dbc0535de9cdde3d57c4a5a57484a88b2e54dc41c4bacc9ede4 +size 7574077 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..24091eace34323a7ecc127e77c656b71bc2efaab --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs7_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs7-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs7 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 7 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed11/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs7/seed_11", + "wandb_unique_id": "demon-attack-fs7-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..d0d79751104d703d6eb460fb4416f3b856624d04 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b187496e645394b6e661c2d4bd3c8e9dd74c3904570c1aae97d8f3c98d20ecc +size 7574077 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..90267234debc7b4dd26a99b1ec61c50bc4b4b31a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs7_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs7-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs7 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 7 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed12/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs7/seed_12", + "wandb_unique_id": "demon-attack-fs7-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..9316e495565cc9119c5596ce8f463fd7c53773fa --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6483c20eb6ba1054fe910201966cf6478a97bf43fd062bdd481d44383ed2635c +size 7574077 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..837d13a70af6c5c4a52fe4f424703ca0fafe5812 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs7_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs7-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs7 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 7 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed13/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs7/seed_13", + "wandb_unique_id": "demon-attack-fs7-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..94e43e2f352e05084a9e37d0ed0ab462d740abc5 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3d9bba134fbe0168cf14e21d3f34e6aadfb63a1b5b3c9203c24975cbe080c7a4 +size 7574077 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..12ac5dbbc17db5597a4f121b529daa7e81b4b00c --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs7:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs7_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs7-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs7 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 7 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs7_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs7-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs7", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 7, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs7_seed14/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs7/seed_14", + "wandb_unique_id": "demon-attack-fs7-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..c6caa1e07e838bb64643734b97864541b76b8774 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:1d54041d3b55ac7ade759c25308890d1a2152c9472c81f717e99a534dc01875f +size 7695165 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..91f3c2984461037e1ce1c00cecabacadb98a05c7 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs8_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs8 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 8 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed10/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs8/seed_10", + "wandb_unique_id": "demon-attack-fs8-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..80ffe544f2e42c6fedf5622ad890e927a62fcf1f --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:db910f310a72eb8045dc6662b97d638f04d7a2b364bf70b962373469fa67f4f0 +size 7695165 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2f1efdd768d03f88c7036caf7c4b7338d00a5e52 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs8_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs8 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 8 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed11/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs8/seed_11", + "wandb_unique_id": "demon-attack-fs8-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..99c3bd84f2c04940ac3718c458e1cf9258782015 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a572bdbd9dbf6bb74c5b947fd05a9b0230118d9649eaef88fd2b5e72e550055f +size 7695165 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..e703d062a801ae035f6dac10cf335d960e159082 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs8_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs8 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 8 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed12/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs8/seed_12", + "wandb_unique_id": "demon-attack-fs8-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..ee13f92083bc768efc0ef57f38c06f43a148a6b7 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:d993111d48305dabaa22cdc3a94c8b17a2f1c4f4a3a6010d7e864ab485608a52 +size 7695165 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..42f6a874d23a9cfd35049230aedc0ae261f19641 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs8_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs8 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 8 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed13/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs8/seed_13", + "wandb_unique_id": "demon-attack-fs8-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..498e3daae9d71ecfb56f3319f2039ef529b61cf9 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3a21a5d805e3776e2630ca9aafbf2ce09817dab07ca81f99483ef9c02d53b0b8 +size 7695165 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..4d0e2cd48df62bd06a787fcf936043e0c3ae717f --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs8:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs8_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 3600 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs8-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs8 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 8 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs8_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs8-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs8", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 8, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 3600, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs8_seed14/episode_metrics.jsonl" + }, + "git_hash": "a81783762c1f965cc4a9d2eb7f028b47ab17a859", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs8/seed_14", + "wandb_unique_id": "demon-attack-fs8-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..b3091f85808bdaa1237ef66e8fe328d7002e135c --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ccd5e6d3204a5f9c51844ff7a2ec743c24571f946e54a77c2b2f04f1f738fa0 +size 7816253 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/config.json new file mode 100644 index 0000000000000000000000000000000000000000..f0231888eccf30cc82aefebdd7691fa8b2852637 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed10/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed10" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed10/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs9_seed10 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 10 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed10/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs9-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs9 seed10 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 9 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed10", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 10, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed10" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed10/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs9/seed_10", + "wandb_unique_id": "demon-attack-fs9-fixed_l4-s10" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..e1c79609d108906c979226bb2adaebf6c70530ae --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2759fde792ffe806a702c006e05ce937d8bc592dc136bce6266d2f222c19f30e +size 7816253 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9b04f77f16e26619da63d60a00da77c270d607e1 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed11/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed11" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed11/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs9_seed11 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 11 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed11/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs9-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs9 seed11 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 9 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed11", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 11, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed11" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed11/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs9/seed_11", + "wandb_unique_id": "demon-attack-fs9-fixed_l4-s11" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..118cea98d7e21b8b27a94f9c90a194f2cd0f54a0 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:fcf204ef1b6dce0fa0e23473ba2b5b26086b6d6e70c2acb533ad19470c74b01d +size 7816253 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/config.json new file mode 100644 index 0000000000000000000000000000000000000000..a144d8c59bcddcaf14ca78468c1660cd8d702b72 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed12/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed12" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed12/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs9_seed12 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 12 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed12/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs9-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs9 seed12 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 9 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed12", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 12, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed12" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed12/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs9/seed_12", + "wandb_unique_id": "demon-attack-fs9-fixed_l4-s12" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..315b240cc0f203544a673a6c286f1a573af5c216 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3bf6ce6620ba04f3d25fd543375fc6e321c1534c08da3ffce4f23f5467217185 +size 7816253 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/config.json new file mode 100644 index 0000000000000000000000000000000000000000..6d7ecd45f372f65de3a39874dc0caacdcfa4f373 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed13/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed13" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed13/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs9_seed13 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 13 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed13/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs9-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs9 seed13 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 9 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed13", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 13, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed13" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed13/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs9/seed_13", + "wandb_unique_id": "demon-attack-fs9-fixed_l4-s13" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..9d4db58df967345ffd8fe74fc10e3b8f829ab0f6 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c43af2706c66442a0a14c6d669eff18bfe364b71265dc4b3d8d417fc27b26480 +size 7816253 diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/config.json b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/config.json new file mode 100644 index 0000000000000000000000000000000000000000..5221146716eb668558504e57c2fd3b8d9d830f0a --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:fixed_l4:fs9:obs15:stride4:seed14/config.json @@ -0,0 +1,253 @@ +{ + "help": false, + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "serial_mode": false, + "batched_sampling": true, + "num_batches_to_accumulate": 2, + "worker_num_splits": 1, + "policy_workers_per_policy": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "shuffle_minibatches": false, + "gamma": 0.99, + "reward_scale": 1.0, + "reward_clip": 1000.0, + "value_bootstrap": false, + "normalize_returns": true, + "exploration_loss_coeff": 0.003, + "value_loss_coeff": 0.5, + "kl_loss_coeff": 0.0, + "exploration_loss": "entropy", + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "with_vtrace": false, + "vtrace_rho": 1.0, + "vtrace_c": 1.0, + "optimizer": "adam", + "adam_eps": 1e-05, + "adam_beta1": 0.9, + "adam_beta2": 0.999, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "lr_schedule": "linear_decay", + "lr_schedule_kl_threshold": 0.008, + "lr_adaptive_min": 1e-06, + "lr_adaptive_max": 0.01, + "obs_subtract_mean": 0.0, + "obs_scale": 255.0, + "normalize_input": true, + "normalize_input_keys": null, + "decorrelate_experience_max_seconds": 0, + "decorrelate_envs_on_one_worker": true, + "actor_worker_gpus": [ + 0 + ], + "set_workers_cpu_affinity": true, + "force_envs_single_thread": false, + "default_niceness": 0, + "log_to_file": true, + "experiment_summaries_interval": 1, + "flush_summaries_interval": 30, + "stats_avg": 100, + "summaries_use_frameskip": true, + "heartbeat_interval": 20, + "heartbeat_reporting_interval": 180, + "train_for_env_steps": 25000000, + "train_for_seconds": 10000000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "load_checkpoint_kind": "latest", + "save_milestones_sec": -1, + "save_best_every_sec": 5, + "save_best_metric": "reward", + "save_best_after": 100000, + "benchmark": false, + "encoder_mlp_layers": [ + 512, + 512 + ], + "encoder_conv_architecture": "convnet_atari", + "encoder_conv_mlp_layers": [ + 512 + ], + "use_rnn": false, + "rnn_size": 512, + "rnn_type": "gru", + "rnn_num_layers": 1, + "decoder_mlp_layers": [], + "nonlinearity": "elu", + "policy_initialization": "orthogonal", + "policy_init_gain": 1.0, + "actor_critic_share_weights": true, + "adaptive_stddev": true, + "continuous_tanh_scale": 0.0, + "initial_stddev": 1.0, + "use_env_info_cache": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "env_frameskip": 1, + "env_framestack": 1, + "pixel_format": "CHW", + "use_record_episode_statistics": false, + "with_wandb": true, + "wandb_user": null, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed14" + ], + "with_pbt": false, + "pbt_mix_policies_in_one_env": true, + "pbt_period_env_steps": 5000000, + "pbt_start_mutation": 20000000, + "pbt_replace_fraction": 0.3, + "pbt_mutation_rate": 0.15, + "pbt_replace_reward_gap": 0.1, + "pbt_replace_reward_gap_absolute": 1e-06, + "pbt_optimize_gamma": false, + "pbt_target_objective": "true_objective", + "pbt_perturb_min": 1.1, + "pbt_perturb_max": 1.5, + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "export_env_raw_rgb_frames": false, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "mean_latency_ms": null, + "std_latency_ms": null, + "min_latency_ms": null, + "max_latency_ms": null, + "latency_seed": null, + "add_latency_info": false, + "max_pending_actions": null, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_latency_raw_frame_values": "0,1,2,3,4,5", + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed14/episode_metrics.jsonl", + "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_fixed_l4_fs9_seed14 --train_dir results/checkpoints_factor_sweeps/demon_attack/context_window --restart_behavior resume --device gpu --actor_worker_gpus 0 --env_gpu_observations True --env_gpu_actions True --gpu-env-batch-size 512 --seed 14 --episode_metrics_path results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed14/episode_metrics.jsonl --train_for_env_steps 25000000 --num_workers 1 --num_envs_per_worker 1 --num_policies 1 --batch_size 1024 --rollout 64 --recurrence 32 --num_epochs 2 --num_batches_per_epoch 8 --worker_num_splits 1 --max_policy_lag 300 --learning_rate 0.000303 --gamma 0.99 --gae_lambda 0.95 --ppo_clip_ratio 0.1 --ppo_clip_value 0.2 --value_loss_coeff 0.5 --max_grad_norm 0.0 --save_every_sec 600 --keep_checkpoints 5 --stats_avg 100 --experiment_summaries_interval 1 --batched_sampling True --async_rl True --batched_sampling True --use_rnn False --normalize_returns True --normalize_input True --latency-type fixed --fixed-latency-ms 66.66666666666667 --add-latency-info False --eval-episodes 100 --eval-parallel-envs 100 --eval-max-steps 36000 --eval-deterministic True --eval-raw-reward --with_wandb True --wandb_project latency-sensitive-bench --wandb_group demon-attack-fs9-fixed_l4 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack fixed fixed_l4 fs9 seed14 --env_id ALE/DemonAttack-v5 --env-fps 60 --obs-fps 15.0 --screen_size 84 --noop_max 30 --frame_stack 9 --terminal_on_life_loss True --clip_reward True --repeat_action_probability 0.0 --full_action_space False --max_episode_steps 25000 --simulator gpu --gpu-env-device auto --hold-policy hold --ordering-policy latest_ready", + "cli_args": { + "algo": "APPO", + "env": "latency_demon_attack", + "experiment": "demon_attack_frame_stack_fixed_l4_fs9_seed14", + "train_dir": "results/checkpoints_factor_sweeps/demon_attack/context_window", + "restart_behavior": "resume", + "device": "gpu", + "seed": 14, + "num_policies": 1, + "async_rl": true, + "batched_sampling": true, + "worker_num_splits": 1, + "max_policy_lag": 300, + "num_workers": 1, + "num_envs_per_worker": 1, + "batch_size": 1024, + "num_batches_per_epoch": 8, + "num_epochs": 2, + "rollout": 64, + "recurrence": 32, + "gamma": 0.99, + "normalize_returns": true, + "value_loss_coeff": 0.5, + "gae_lambda": 0.95, + "ppo_clip_ratio": 0.1, + "ppo_clip_value": 0.2, + "max_grad_norm": 0.0, + "learning_rate": 0.000303, + "normalize_input": true, + "actor_worker_gpus": [ + 0 + ], + "experiment_summaries_interval": 1, + "stats_avg": 100, + "train_for_env_steps": 25000000, + "save_every_sec": 600, + "keep_checkpoints": 5, + "use_rnn": false, + "env_gpu_actions": true, + "env_gpu_observations": true, + "with_wandb": true, + "wandb_project": "latency-sensitive-bench", + "wandb_group": "demon-attack-fs9-fixed_l4", + "wandb_job_type": "sample_factory", + "wandb_tags": [ + "factor_sweep", + "demon-attack", + "frame_stack", + "fixed", + "fixed_l4", + "fs9", + "seed14" + ], + "env_id": "ALE/DemonAttack-v5", + "env_fps": 60.0, + "obs_fps": 15.0, + "screen_size": 84, + "noop_max": 30, + "frame_stack": 9, + "terminal_on_life_loss": true, + "clip_reward": true, + "repeat_action_probability": 0.0, + "full_action_space": false, + "max_episode_steps": 25000, + "simulator": "gpu", + "gpu_env_device": "auto", + "gpu_env_batch_size": 512, + "mode": "train", + "latency_type": "fixed", + "fixed_latency_ms": 66.66666666666667, + "add_latency_info": false, + "hold_policy": "hold", + "ordering_policy": "latest_ready", + "eval_episodes": 100, + "eval_parallel_envs": 100, + "eval_max_steps": 36000, + "eval_deterministic": true, + "eval_raw_reward": true, + "episode_metrics_path": "results/checkpoints_factor_sweeps/demon_attack/context_window/demon_attack_frame_stack_fixed_l4_fs9_seed14/episode_metrics.jsonl" + }, + "git_hash": "eb3a2e1efbd2aa03a60d7f44f5e18d8fdd0f5a2d", + "git_repo_name": "git@github.com:ZihanWang314/latency-sensitive-bench.git", + "eval_env_frameskip": 1, + "output_dir": "outputs/factor_sweeps/demon_attack/context_window/train/frame_stack/fixed_l4/fs9/seed_14", + "wandb_unique_id": "demon-attack-fs9-fixed_l4-s14" +} \ No newline at end of file diff --git a/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth new file mode 100644 index 0000000000000000000000000000000000000000..f5ec3816d82f65dbc280d8e66e0e051af3f11e68 --- /dev/null +++ b/latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c4d27b3ccb39c54cd8e4dfae1656743233cd709a47f7035cb39e424b7fa98173 +size 7937341