Upload config.json with huggingface_hub
Browse files- config.json +141 -0
config.json
ADDED
@@ -0,0 +1,141 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"help": false,
|
3 |
+
"algo": "APPO",
|
4 |
+
"env": "doom_health_gathering_supreme",
|
5 |
+
"experiment": "default_experiment",
|
6 |
+
"train_dir": "/home/micha/notes/topics/rl/train_dir",
|
7 |
+
"restart_behavior": "resume",
|
8 |
+
"device": "gpu",
|
9 |
+
"seed": null,
|
10 |
+
"num_policies": 1,
|
11 |
+
"async_rl": true,
|
12 |
+
"serial_mode": false,
|
13 |
+
"batched_sampling": false,
|
14 |
+
"num_batches_to_accumulate": 2,
|
15 |
+
"worker_num_splits": 2,
|
16 |
+
"policy_workers_per_policy": 1,
|
17 |
+
"max_policy_lag": 1000,
|
18 |
+
"num_workers": 8,
|
19 |
+
"num_envs_per_worker": 4,
|
20 |
+
"batch_size": 1024,
|
21 |
+
"num_batches_per_epoch": 1,
|
22 |
+
"num_epochs": 1,
|
23 |
+
"rollout": 32,
|
24 |
+
"recurrence": 32,
|
25 |
+
"shuffle_minibatches": false,
|
26 |
+
"gamma": 0.99,
|
27 |
+
"reward_scale": 1.0,
|
28 |
+
"reward_clip": 1000.0,
|
29 |
+
"value_bootstrap": false,
|
30 |
+
"normalize_returns": true,
|
31 |
+
"exploration_loss_coeff": 0.001,
|
32 |
+
"value_loss_coeff": 0.5,
|
33 |
+
"kl_loss_coeff": 0.0,
|
34 |
+
"exploration_loss": "symmetric_kl",
|
35 |
+
"gae_lambda": 0.95,
|
36 |
+
"ppo_clip_ratio": 0.1,
|
37 |
+
"ppo_clip_value": 0.2,
|
38 |
+
"with_vtrace": false,
|
39 |
+
"vtrace_rho": 1.0,
|
40 |
+
"vtrace_c": 1.0,
|
41 |
+
"optimizer": "adam",
|
42 |
+
"adam_eps": 1e-06,
|
43 |
+
"adam_beta1": 0.9,
|
44 |
+
"adam_beta2": 0.999,
|
45 |
+
"max_grad_norm": 4.0,
|
46 |
+
"learning_rate": 0.0001,
|
47 |
+
"lr_schedule": "constant",
|
48 |
+
"lr_schedule_kl_threshold": 0.008,
|
49 |
+
"obs_subtract_mean": 0.0,
|
50 |
+
"obs_scale": 255.0,
|
51 |
+
"normalize_input": true,
|
52 |
+
"normalize_input_keys": null,
|
53 |
+
"decorrelate_experience_max_seconds": 0,
|
54 |
+
"decorrelate_envs_on_one_worker": true,
|
55 |
+
"actor_worker_gpus": [],
|
56 |
+
"set_workers_cpu_affinity": true,
|
57 |
+
"force_envs_single_thread": false,
|
58 |
+
"default_niceness": 0,
|
59 |
+
"log_to_file": true,
|
60 |
+
"experiment_summaries_interval": 10,
|
61 |
+
"flush_summaries_interval": 30,
|
62 |
+
"stats_avg": 100,
|
63 |
+
"summaries_use_frameskip": true,
|
64 |
+
"heartbeat_interval": 20,
|
65 |
+
"heartbeat_reporting_interval": 600,
|
66 |
+
"train_for_env_steps": 4000000,
|
67 |
+
"train_for_seconds": 10000000000,
|
68 |
+
"save_every_sec": 120,
|
69 |
+
"keep_checkpoints": 2,
|
70 |
+
"load_checkpoint_kind": "latest",
|
71 |
+
"save_milestones_sec": -1,
|
72 |
+
"save_best_every_sec": 5,
|
73 |
+
"save_best_metric": "reward",
|
74 |
+
"save_best_after": 100000,
|
75 |
+
"benchmark": false,
|
76 |
+
"encoder_mlp_layers": [
|
77 |
+
512,
|
78 |
+
512
|
79 |
+
],
|
80 |
+
"encoder_conv_architecture": "convnet_simple",
|
81 |
+
"encoder_conv_mlp_layers": [
|
82 |
+
512
|
83 |
+
],
|
84 |
+
"use_rnn": true,
|
85 |
+
"rnn_size": 512,
|
86 |
+
"rnn_type": "gru",
|
87 |
+
"rnn_num_layers": 1,
|
88 |
+
"decoder_mlp_layers": [],
|
89 |
+
"nonlinearity": "elu",
|
90 |
+
"policy_initialization": "orthogonal",
|
91 |
+
"policy_init_gain": 1.0,
|
92 |
+
"actor_critic_share_weights": true,
|
93 |
+
"adaptive_stddev": true,
|
94 |
+
"continuous_tanh_scale": 0.0,
|
95 |
+
"initial_stddev": 1.0,
|
96 |
+
"use_env_info_cache": false,
|
97 |
+
"env_gpu_actions": false,
|
98 |
+
"env_gpu_observations": true,
|
99 |
+
"env_frameskip": 4,
|
100 |
+
"env_framestack": 1,
|
101 |
+
"pixel_format": "CHW",
|
102 |
+
"use_record_episode_statistics": false,
|
103 |
+
"with_wandb": false,
|
104 |
+
"wandb_user": null,
|
105 |
+
"wandb_project": "sample_factory",
|
106 |
+
"wandb_group": null,
|
107 |
+
"wandb_job_type": "SF",
|
108 |
+
"wandb_tags": [],
|
109 |
+
"with_pbt": false,
|
110 |
+
"pbt_mix_policies_in_one_env": true,
|
111 |
+
"pbt_period_env_steps": 5000000,
|
112 |
+
"pbt_start_mutation": 20000000,
|
113 |
+
"pbt_replace_fraction": 0.3,
|
114 |
+
"pbt_mutation_rate": 0.15,
|
115 |
+
"pbt_replace_reward_gap": 0.1,
|
116 |
+
"pbt_replace_reward_gap_absolute": 1e-06,
|
117 |
+
"pbt_optimize_gamma": false,
|
118 |
+
"pbt_target_objective": "true_objective",
|
119 |
+
"pbt_perturb_min": 1.1,
|
120 |
+
"pbt_perturb_max": 1.5,
|
121 |
+
"num_agents": -1,
|
122 |
+
"num_humans": 0,
|
123 |
+
"num_bots": -1,
|
124 |
+
"start_bot_difficulty": null,
|
125 |
+
"timelimit": null,
|
126 |
+
"res_w": 128,
|
127 |
+
"res_h": 72,
|
128 |
+
"wide_aspect_ratio": false,
|
129 |
+
"eval_env_frameskip": 1,
|
130 |
+
"fps": 35,
|
131 |
+
"command_line": "--env=doom_health_gathering_supreme --num_workers=8 --num_envs_per_worker=4 --train_for_env_steps=4000000",
|
132 |
+
"cli_args": {
|
133 |
+
"env": "doom_health_gathering_supreme",
|
134 |
+
"num_workers": 8,
|
135 |
+
"num_envs_per_worker": 4,
|
136 |
+
"train_for_env_steps": 4000000
|
137 |
+
},
|
138 |
+
"git_hash": "fd8c95f7335d659f45f7b4156c346717dbce2417",
|
139 |
+
"git_repo_name": "https://github.com/oxidase/home",
|
140 |
+
"train_script": "unit8_doom"
|
141 |
+
}
|