Import factor-sweep assets 1-400

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs10_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs10/seed_10",
+  "wandb_unique_id": "demon-attack-fs10-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7e704d2345337c461e445ee123e1b6b44816a392b27a44de523c2c484b13262
+size 7937341

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed11/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 11,
+  "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_uniform_u2_6_fs10_seed11/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 11 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 11,
+    "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_uniform_u2_6_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/uniform_u2_6/fs10/seed_11",
+  "wandb_unique_id": "demon-attack-fs10-uniform_u2_6-s11"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa3ba64b1c0004383736d0ab9053c93331c6c5acbdf2d77f921be40e4769d294
+size 7937341

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed12/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 12,
+  "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_uniform_u2_6_fs10_seed12/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 12 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 12,
+    "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_uniform_u2_6_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/uniform_u2_6/fs10/seed_12",
+  "wandb_unique_id": "demon-attack-fs10-uniform_u2_6-s12"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64a4f25efe6ac43fc878f2b6787da671a7b573c90992acbefe03d7b91283e62c
+size 7937341

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed13/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 13,
+  "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_uniform_u2_6_fs10_seed13/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 13 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 13,
+    "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_uniform_u2_6_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/uniform_u2_6/fs10/seed_13",
+  "wandb_unique_id": "demon-attack-fs10-uniform_u2_6-s13"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e0cbb7fff9c3b356151914f565d0e488601604c2d476076b98e386b8d5b358c
+size 7937341

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs10:obs15:stride4:seed14/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 14,
+  "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_uniform_u2_6_fs10_seed14/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 14 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 14,
+    "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_uniform_u2_6_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/uniform_u2_6/fs10/seed_14",
+  "wandb_unique_id": "demon-attack-fs10-uniform_u2_6-s14"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7ea434aa8641a99366952230b35aac7cacf187d9a1ca0a01d0ad8685760ff14
+size 8058429

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs11_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs11/seed_10",
+  "wandb_unique_id": "demon-attack-fs11-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31cf534081733e58389ecfab7c9d6d1d3e9958a14bcf68018b0ae2681cf4333e
+size 8058429

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed11/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 11,
+  "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_uniform_u2_6_fs11_seed11/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 11 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 11,
+    "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_uniform_u2_6_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/uniform_u2_6/fs11/seed_11",
+  "wandb_unique_id": "demon-attack-fs11-uniform_u2_6-s11"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed5f20466a803db45b8efbc4b5774f7b9897a5859fa07dabb1e632afc0931ffe
+size 8058429

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed12/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 12,
+  "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_uniform_u2_6_fs11_seed12/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 12 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 12,
+    "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_uniform_u2_6_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/uniform_u2_6/fs11/seed_12",
+  "wandb_unique_id": "demon-attack-fs11-uniform_u2_6-s12"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e37c8d58bf41b2143f2f66a4cd7f4af5fd71fdb4fb78990c27150b2d68af315b
+size 8058429

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed13/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 13,
+  "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_uniform_u2_6_fs11_seed13/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 13 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 13,
+    "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_uniform_u2_6_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/uniform_u2_6/fs11/seed_13",
+  "wandb_unique_id": "demon-attack-fs11-uniform_u2_6-s13"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b643c5cfee6c062ace24c8fc4904d330ce958b659005c6125a565ec5d8d55291
+size 8058429

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs11:obs15:stride4:seed14/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 14,
+  "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_uniform_u2_6_fs11_seed14/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 14 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 14,
+    "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_uniform_u2_6_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/uniform_u2_6/fs11/seed_14",
+  "wandb_unique_id": "demon-attack-fs11-uniform_u2_6-s14"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b3a4e32d55a89eccfe74620006f04f5b8419640d1b6b0ec43a513af9042b906
+size 8179517

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs12_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs12/seed_10",
+  "wandb_unique_id": "demon-attack-fs12-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77fae23d456b496f72c322d935af05d0c86614fa393c94d40063fc15b28aa70b
+size 8179517

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed11/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 11,
+  "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_uniform_u2_6_fs12_seed11/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 11 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 11,
+    "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_uniform_u2_6_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/uniform_u2_6/fs12/seed_11",
+  "wandb_unique_id": "demon-attack-fs12-uniform_u2_6-s11"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27554d3be4fca11bd98ab6e9ebceea30222c8fff55ce68aae64d2a41c00a9d29
+size 8179517

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed12/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 12,
+  "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_uniform_u2_6_fs12_seed12/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 12 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 12,
+    "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_uniform_u2_6_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/uniform_u2_6/fs12/seed_12",
+  "wandb_unique_id": "demon-attack-fs12-uniform_u2_6-s12"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b10e00b1a92f6d81791d76e4781830348fde29c7a863f7f0516cd94a27b2c2e
+size 8179517

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs12:obs15:stride4:seed13/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 13,
+  "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_uniform_u2_6_fs12_seed13/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 13 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 13,
+    "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_uniform_u2_6_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/uniform_u2_6/fs12/seed_13",
+  "wandb_unique_id": "demon-attack-fs12-uniform_u2_6-s13"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89d652bc51e3ac31ddf21294c5b9b5222e27b3c5de40529cd6afaa575b910b2f
+size 7210813

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs4_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs4/seed_10",
+  "wandb_unique_id": "demon-attack-fs4-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f185ff986491332f17c8f620166d706f12fd46519ec0b4ff410b84cb055f07a9
+size 7210813

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed11/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 11,
+  "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_uniform_u2_6_fs4_seed11/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 11 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 11,
+    "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_uniform_u2_6_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/uniform_u2_6/fs4/seed_11",
+  "wandb_unique_id": "demon-attack-fs4-uniform_u2_6-s11"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d69759fbea359b854109721ba87ae3f5e7812fff3c2371c7f92f495caa30a2e
+size 7210813

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed12/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 12,
+  "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_uniform_u2_6_fs4_seed12/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 12 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 12,
+    "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_uniform_u2_6_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/uniform_u2_6/fs4/seed_12",
+  "wandb_unique_id": "demon-attack-fs4-uniform_u2_6-s12"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:213f4e418d3c60d62e9fbf67094607d43e53e5af93090ecb3bdefe984ce1b3da
+size 7210813

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed13/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 13,
+  "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_uniform_u2_6_fs4_seed13/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 13 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 13,
+    "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_uniform_u2_6_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/uniform_u2_6/fs4/seed_13",
+  "wandb_unique_id": "demon-attack-fs4-uniform_u2_6-s13"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3052199dadf157e45685e7d170eb8cd2b0de5792a80a62bc4228372053d82e7e
+size 7210813

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs4:obs15:stride4:seed14/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 14,
+  "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_uniform_u2_6_fs4_seed14/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 14 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 14,
+    "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_uniform_u2_6_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/uniform_u2_6/fs4/seed_14",
+  "wandb_unique_id": "demon-attack-fs4-uniform_u2_6-s14"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67c5cf14508833f5106331ebf2f2cbc54f802d17bb7903e2d865ddb23ac34b2a
+size 7331901

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs5_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs5/seed_10",
+  "wandb_unique_id": "demon-attack-fs5-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47597f966b12da8f4e8720db6045f5ec1716672e93667400d7a508f39fcc9c2e
+size 7331901

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed11/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 11,
+  "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_uniform_u2_6_fs5_seed11/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 11 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 11,
+    "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_uniform_u2_6_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/uniform_u2_6/fs5/seed_11",
+  "wandb_unique_id": "demon-attack-fs5-uniform_u2_6-s11"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed12/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6041f7be23583fc1d6a37d253b5c45de1e2008f4ee88a6aa1271a243ca90859
+size 7331901

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed12/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 12,
+  "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_uniform_u2_6_fs5_seed12/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 12 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 12,
+    "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_uniform_u2_6_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/uniform_u2_6/fs5/seed_12",
+  "wandb_unique_id": "demon-attack-fs5-uniform_u2_6-s12"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed13/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56d7c5bf2bf28d163a2f7fb20f34c42d3caedcebd3e0e604821742eadafcf606
+size 7331901

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed13/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 13,
+  "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_uniform_u2_6_fs5_seed13/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 13 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 13,
+    "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_uniform_u2_6_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/uniform_u2_6/fs5/seed_13",
+  "wandb_unique_id": "demon-attack-fs5-uniform_u2_6-s13"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed14/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ff7f9b130d633f55a642ede33485441297ea6b7b763546d0f3edd89a2d409dd
+size 7331901

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs5:obs15:stride4:seed14/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "frame_stack",
+    "uniform",
+    "uniform_u2_6",
+    "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 14,
+  "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_uniform_u2_6_fs5_seed14/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 14 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 14,
+    "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_uniform_u2_6_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/uniform_u2_6/fs5/seed_14",
+  "wandb_unique_id": "demon-attack-fs5-uniform_u2_6-s14"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs6:obs15:stride4:seed10/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b317e768314d95b0e485def34bee8e44c385a46e4d862d206741e3f1b8a80be
+size 7452989

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs6:obs15:stride4:seed10/config.json

@@ -0,0 +1,255 @@
+{
+  "help": false,
+  "algo": "APPO",
+  "env": "latency_demon_attack",
+  "experiment": "demon_attack_frame_stack_uniform_u2_6_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-uniform_u2_6",
+  "wandb_job_type": "sample_factory",
+  "wandb_tags": [
+    "factor_sweep",
+    "demon-attack",
+    "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,
+  "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": "uniform_distribution",
+  "fixed_latency_ms": null,
+  "mean_latency_ms": null,
+  "std_latency_ms": null,
+  "min_latency_ms": 33.333333333333336,
+  "max_latency_ms": 100.0,
+  "latency_seed": 10,
+  "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_uniform_u2_6_fs6_seed10/episode_metrics.jsonl",
+  "command_line": "--mode train --algo APPO --env latency_demon_attack --experiment demon_attack_frame_stack_uniform_u2_6_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_uniform_u2_6_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 uniform_distribution --min-latency-ms 33.333333333333336 --max-latency-ms 100.0 --latency-seed 10 --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-uniform_u2_6 --wandb_job_type sample_factory --wandb_tags factor_sweep demon-attack frame_stack uniform uniform_u2_6 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_uniform_u2_6_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-uniform_u2_6",
+    "wandb_job_type": "sample_factory",
+    "wandb_tags": [
+      "factor_sweep",
+      "demon-attack",
+      "frame_stack",
+      "uniform",
+      "uniform_u2_6",
+      "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": "uniform_distribution",
+    "min_latency_ms": 33.333333333333336,
+    "max_latency_ms": 100.0,
+    "latency_seed": 10,
+    "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_uniform_u2_6_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/uniform_u2_6/fs6/seed_10",
+  "wandb_unique_id": "demon-attack-fs6-uniform_u2_6-s10"
+}

latency-aware/demon-attack/small-policy/factor_sweep:demon_attack:frame_stack:uniform_u2_6:fs6:obs15:stride4:seed11/checkpoint_p0/checkpoint_000048880_25026560.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a2b5c018f7add7f106beb2941ecfe12f2bf0322f0aebb3b604ea300c87801f5
+size 7452989