Commit
•
d295ca2
1
Parent(s):
d84e2e9
Upload cfg.json with huggingface_hub
Browse files
cfg.json
ADDED
@@ -0,0 +1,212 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"help": false,
|
3 |
+
"algo": "APPO",
|
4 |
+
"env": "quadrotor_multi",
|
5 |
+
"experiment": "01_baseline_see_1111",
|
6 |
+
"train_dir": "./train_dir/quad_single_baseline/baseline_slurm_norm",
|
7 |
+
"restart_behavior": "resume",
|
8 |
+
"device": "gpu",
|
9 |
+
"seed": 1111,
|
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": 100000000,
|
18 |
+
"num_workers": 36,
|
19 |
+
"num_envs_per_worker": 4,
|
20 |
+
"batch_size": 1024,
|
21 |
+
"num_batches_per_epoch": 1,
|
22 |
+
"num_epochs": 1,
|
23 |
+
"rollout": 128,
|
24 |
+
"recurrence": 1,
|
25 |
+
"shuffle_minibatches": true,
|
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": 0.0,
|
32 |
+
"value_loss_coeff": 0.5,
|
33 |
+
"kl_loss_coeff": 0.0,
|
34 |
+
"exploration_loss": "entropy",
|
35 |
+
"gae_lambda": 1.0,
|
36 |
+
"ppo_clip_ratio": 0.1,
|
37 |
+
"ppo_clip_value": 5.0,
|
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": 5.0,
|
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": 1.0,
|
51 |
+
"normalize_input": true,
|
52 |
+
"decorrelate_experience_max_seconds": 10,
|
53 |
+
"decorrelate_envs_on_one_worker": true,
|
54 |
+
"actor_worker_gpus": [],
|
55 |
+
"set_workers_cpu_affinity": true,
|
56 |
+
"force_envs_single_thread": true,
|
57 |
+
"default_niceness": 0,
|
58 |
+
"experiment_summaries_interval": 60,
|
59 |
+
"stats_avg": 100,
|
60 |
+
"train_for_env_steps": 1000000000,
|
61 |
+
"train_for_seconds": 10000000000,
|
62 |
+
"save_every_sec": 120,
|
63 |
+
"keep_checkpoints": 3,
|
64 |
+
"load_checkpoint_kind": "latest",
|
65 |
+
"save_milestones_sec": 10000,
|
66 |
+
"save_best_every_sec": 5,
|
67 |
+
"save_best_metric": "reward",
|
68 |
+
"save_best_after": 100000,
|
69 |
+
"benchmark": false,
|
70 |
+
"encoder_type": "mlp",
|
71 |
+
"encoder_subtype": "mlp_quads",
|
72 |
+
"encoder_custom": "quad_multi_encoder",
|
73 |
+
"encoder_extra_fc_layers": 0,
|
74 |
+
"hidden_size": 16,
|
75 |
+
"nonlinearity": "tanh",
|
76 |
+
"policy_initialization": "xavier_uniform",
|
77 |
+
"policy_init_gain": 1.0,
|
78 |
+
"actor_critic_share_weights": false,
|
79 |
+
"adaptive_stddev": false,
|
80 |
+
"initial_stddev": 1.0,
|
81 |
+
"use_rnn": false,
|
82 |
+
"rnn_type": "gru",
|
83 |
+
"rnn_num_layers": 1,
|
84 |
+
"env_gpu_actions": false,
|
85 |
+
"env_frameskip": 1,
|
86 |
+
"env_framestack": 4,
|
87 |
+
"pixel_format": "CHW",
|
88 |
+
"with_wandb": true,
|
89 |
+
"wandb_user": null,
|
90 |
+
"wandb_project": "sample_factory",
|
91 |
+
"wandb_group": null,
|
92 |
+
"wandb_job_type": "SF",
|
93 |
+
"wandb_tags": [
|
94 |
+
"sf2",
|
95 |
+
"normalize"
|
96 |
+
],
|
97 |
+
"with_pbt": false,
|
98 |
+
"pbt_mix_policies_in_one_env": true,
|
99 |
+
"pbt_period_env_steps": 5000000,
|
100 |
+
"pbt_start_mutation": 20000000,
|
101 |
+
"pbt_replace_fraction": 0.3,
|
102 |
+
"pbt_mutation_rate": 0.15,
|
103 |
+
"pbt_replace_reward_gap": 0.1,
|
104 |
+
"pbt_replace_reward_gap_absolute": 1e-06,
|
105 |
+
"pbt_optimize_batch_size": false,
|
106 |
+
"pbt_optimize_gamma": false,
|
107 |
+
"pbt_target_objective": "true_objective",
|
108 |
+
"pbt_perturb_min": 1.05,
|
109 |
+
"pbt_perturb_max": 1.5,
|
110 |
+
"quads_discretize_actions": -1,
|
111 |
+
"quads_clip_input": false,
|
112 |
+
"quads_effort_reward": null,
|
113 |
+
"quads_episode_duration": 15.0,
|
114 |
+
"quads_num_agents": 1,
|
115 |
+
"quads_neighbor_hidden_size": 256,
|
116 |
+
"quads_neighbor_encoder_type": "no_encoder",
|
117 |
+
"quads_collision_reward": 5.0,
|
118 |
+
"quads_collision_obstacle_reward": 0.0,
|
119 |
+
"quads_settle_reward": 0.0,
|
120 |
+
"quads_settle": false,
|
121 |
+
"quads_vel_reward_out_range": 0.8,
|
122 |
+
"quads_settle_range_meters": 1.0,
|
123 |
+
"quads_collision_hitbox_radius": 2.0,
|
124 |
+
"quads_collision_falloff_radius": 4.0,
|
125 |
+
"quads_collision_smooth_max_penalty": 10.0,
|
126 |
+
"neighbor_obs_type": "none",
|
127 |
+
"quads_use_numba": true,
|
128 |
+
"quads_obstacle_mode": "no_obstacles",
|
129 |
+
"quads_obstacle_num": 0,
|
130 |
+
"quads_obstacle_type": "sphere",
|
131 |
+
"quads_obstacle_size": 0.0,
|
132 |
+
"quads_obstacle_traj": "gravity",
|
133 |
+
"quads_local_obs": -1,
|
134 |
+
"quads_local_coeff": 1.0,
|
135 |
+
"quads_local_metric": "dist",
|
136 |
+
"quads_view_mode": "local",
|
137 |
+
"quads_adaptive_env": false,
|
138 |
+
"quads_mode": "mix",
|
139 |
+
"quads_formation": "circle_horizontal",
|
140 |
+
"quads_formation_size": 0.0,
|
141 |
+
"room_dims": [
|
142 |
+
10,
|
143 |
+
10,
|
144 |
+
10
|
145 |
+
],
|
146 |
+
"quads_obs_repr": "xyz_vxyz_R_omega",
|
147 |
+
"replay_buffer_sample_prob": 0.0,
|
148 |
+
"anneal_collision_steps": 0.0,
|
149 |
+
"quads_obstacle_obs_mode": "relative",
|
150 |
+
"quads_obstacle_hidden_size": 32,
|
151 |
+
"quads_collision_obst_smooth_max_penalty": 10.0,
|
152 |
+
"quads_obst_penalty_fall_off": 10.0,
|
153 |
+
"use_spectral_norm": false,
|
154 |
+
"command_line": "--env=quadrotor_multi --train_for_env_steps=1000000000 --algo=APPO --use_rnn=False --num_workers=36 --num_envs_per_worker=4 --learning_rate=0.0001 --ppo_clip_value=5.0 --recurrence=1 --nonlinearity=tanh --actor_critic_share_weights=False --policy_initialization=xavier_uniform --adaptive_stddev=False --with_vtrace=False --max_policy_lag=100000000 --hidden_size=256 --gae_lambda=1.00 --max_grad_norm=5.0 --exploration_loss_coeff=0.0 --rollout=128 --batch_size=1024 --quads_use_numba=True --quads_mode=mix --quads_episode_duration=15.0 --quads_formation_size=0.0 --encoder_custom=quad_multi_encoder --quads_collision_reward=5.0 --quads_neighbor_hidden_size=256 --neighbor_obs_type=pos_vel --quads_settle_reward=0.0 --quads_collision_hitbox_radius=2.0 --quads_collision_falloff_radius=4.0 --quads_local_obs=6 --quads_local_metric=dist --quads_local_coeff=1.0 --quads_num_agents=8 --quads_collision_reward=5.0 --quads_collision_smooth_max_penalty=10.0 --quads_neighbor_encoder_type=attention --replay_buffer_sample_prob=0.75 --anneal_collision_steps=300000000 --train_for_env_steps=1000000000 --hidden_size=16 --neighbor_obs_type=none --quads_local_obs=-1 --quads_num_agents=1 --replay_buffer_sample_prob=0.0 --anneal_collision_steps=0 --save_milestones_sec=10000 --normalize_input=True --normalize_returns=True --quads_neighbor_encoder_type=no_encoder --serial_mode=False --with_wandb=True --wandb_tags sf2 normalize --seed=1111 --experiment=01_baseline_see_1111 --train_dir=./train_dir/quad_single_baseline/baseline_slurm_norm",
|
155 |
+
"cli_args": {
|
156 |
+
"algo": "APPO",
|
157 |
+
"env": "quadrotor_multi",
|
158 |
+
"experiment": "01_baseline_see_1111",
|
159 |
+
"train_dir": "./train_dir/quad_single_baseline/baseline_slurm_norm",
|
160 |
+
"seed": 1111,
|
161 |
+
"serial_mode": false,
|
162 |
+
"max_policy_lag": 100000000,
|
163 |
+
"num_workers": 36,
|
164 |
+
"num_envs_per_worker": 4,
|
165 |
+
"batch_size": 1024,
|
166 |
+
"rollout": 128,
|
167 |
+
"recurrence": 1,
|
168 |
+
"normalize_returns": true,
|
169 |
+
"exploration_loss_coeff": 0.0,
|
170 |
+
"gae_lambda": 1.0,
|
171 |
+
"ppo_clip_value": 5.0,
|
172 |
+
"with_vtrace": false,
|
173 |
+
"max_grad_norm": 5.0,
|
174 |
+
"learning_rate": 0.0001,
|
175 |
+
"normalize_input": true,
|
176 |
+
"train_for_env_steps": 1000000000,
|
177 |
+
"save_milestones_sec": 10000,
|
178 |
+
"encoder_custom": "quad_multi_encoder",
|
179 |
+
"hidden_size": 16,
|
180 |
+
"nonlinearity": "tanh",
|
181 |
+
"policy_initialization": "xavier_uniform",
|
182 |
+
"actor_critic_share_weights": false,
|
183 |
+
"adaptive_stddev": false,
|
184 |
+
"use_rnn": false,
|
185 |
+
"with_wandb": true,
|
186 |
+
"wandb_tags": [
|
187 |
+
"sf2",
|
188 |
+
"normalize"
|
189 |
+
],
|
190 |
+
"quads_episode_duration": 15.0,
|
191 |
+
"quads_num_agents": 1,
|
192 |
+
"quads_neighbor_hidden_size": 256,
|
193 |
+
"quads_neighbor_encoder_type": "no_encoder",
|
194 |
+
"quads_collision_reward": 5.0,
|
195 |
+
"quads_settle_reward": 0.0,
|
196 |
+
"quads_collision_hitbox_radius": 2.0,
|
197 |
+
"quads_collision_falloff_radius": 4.0,
|
198 |
+
"quads_collision_smooth_max_penalty": 10.0,
|
199 |
+
"neighbor_obs_type": "none",
|
200 |
+
"quads_use_numba": true,
|
201 |
+
"quads_local_obs": -1,
|
202 |
+
"quads_local_coeff": 1.0,
|
203 |
+
"quads_local_metric": "dist",
|
204 |
+
"quads_mode": "mix",
|
205 |
+
"quads_formation_size": 0.0,
|
206 |
+
"replay_buffer_sample_prob": 0.0,
|
207 |
+
"anneal_collision_steps": 0.0
|
208 |
+
},
|
209 |
+
"git_hash": "f6d8f865deb8fe4c36db1e96f4e0fb6220f628b1",
|
210 |
+
"git_repo_name": "https://github.com/andrewzhang505/sample-factory.git",
|
211 |
+
"wandb_unique_id": "01_baseline_see_1111_20220720_104137_983199"
|
212 |
+
}
|