ReBRAC, V-D4RL: Evaluation for Best Hyperparameters
Table 5 from the paper
Created on May 14|Last edited on May 15
Comment
Walker Walk
eval/normalized_score
eval/normalized_score
dataset_name: medium_replay, actor_bc_coef: 0.3, critic_bc_coef: 0.01
dataset_name: random, actor_bc_coef: 0.03, critic_bc_coef: 0.1
dataset_name: medium, actor_bc_coef: 0.03, critic_bc_coef: 0.005
dataset_name: medium_expert, actor_bc_coef: 0.3, critic_bc_coef: 0.005
dataset_name: expert, actor_bc_coef: 0.1, critic_bc_coef: 0.01
Run set
8358
Name
25 visualized
dataset_name: random
dataset_name: random
1
5
dataset_name: medium_replay
dataset_name: medium_replay
1
5
dataset_name: medium
dataset_name: medium
1
5
dataset_name: medium_expert
dataset_name: medium_expert
1
5
dataset_name: expert
dataset_name: expert
1
5
State
Notes
User
Tags
Created
Runtime
Sweep
actor_bc_coef
actor_learning_rate
actor_ln
actor_n_hiddens
batch_size
bc_coef
bc_coef_mul
buffer_size
config_path
critic_bc_coef
critic_learning_rate
critic_ln
critic_n_hiddens
dataset_name
device
discount
encoder_learning_rate
eval_episodes
eval_every
eval_freq
eval_seed
expl_noise
gamma
group
hidden_dim
load_model
max_timesteps
min_decay_coef
mixing_ratio
mlc_job_name
n_episodes
name
noise_clip
normalize
normalize_q
normalize_reward
normalize_states
num_epochs
num_offline_updates
num_online_updates
num_updates_on_epoch
num_warmup_steps
policy_freq
policy_noise
Finished
adagrad
9h 15m 28s
-
0.03
0.0003
-
-
256
-
-
1000000
configs/sac-doup/walker_walk/random.yaml
0.1
0.0003
true
-
random
cuda
0.99
0.0003
-
-
1000
0
0.1
-
rebrac-walker_walk-random-v2-sweep-v0
256
300000
-
-
-
10
["ReBRAC-walker_walk-random-44b25725","ReBRAC-walker_walk-random-7721442e","ReBRAC-walker_walk-random-7fd5164a","ReBRAC-walker_walk-random-838754f7","ReBRAC-walker_walk-random-be0c2d3f"]
0.5
false
-
false
-
-
-
-
-
-
2
0.2
Finished
adagrad
10h 24m
-
0.3
0.0003
-
-
256
-
-
1000000
configs/sac-doup/walker_walk/medium_replay.yaml
0.01
0.0003
true
-
medium_replay
cuda
0.99
0.0003
-
-
1000
0
0.1
-
rebrac-walker_walk-medium-replay-v2-sweep-v0
256
300000
-
-
-
10
["ReBRAC-walker_walk-medium_replay-0917d262","ReBRAC-walker_walk-medium_replay-651aca86","ReBRAC-walker_walk-medium_replay-76469904","ReBRAC-walker_walk-medium_replay-cdbafdad","ReBRAC-walker_walk-medium_replay-ec0da338"]
0.5
false
-
false
-
-
-
-
-
-
2
0.2
Finished
adagrad
9h 56m 35s
-
0.03
0.0003
-
-
256
-
-
1000000
configs/sac-doup/walker_walk/medium.yaml
0.005
0.0003
true
-
medium
cuda
0.99
0.0003
-
-
1000
0
0.1
-
rebrac-walker_walk-medium-v2-sweep-v0
256
300000
-
-
-
10
["ReBRAC-walker_walk-medium-06309a0c","ReBRAC-walker_walk-medium-215aac7c","ReBRAC-walker_walk-medium-77d2cd0d","ReBRAC-walker_walk-medium-9ca37727","ReBRAC-walker_walk-medium-a667d7f1"]
0.5
false
-
false
-
-
-
-
-
-
2
0.2
Finished
adagrad
8h 14m 52s
-
0.3
0.0003
-
-
256
-
-
1000000
configs/sac-doup/walker_walk/medium_expert.yaml
0.005
0.0003
true
-
medium_expert
cuda
0.99
0.0003
-
-
1000
0
0.1
-
rebrac-walker_walk-medium-expert-v2-sweep-v0
256
300000
-
-
-
10
["ReBRAC-walker_walk-medium_expert-0e459435","ReBRAC-walker_walk-medium_expert-1deb9b36","ReBRAC-walker_walk-medium_expert-276c4a8b","ReBRAC-walker_walk-medium_expert-a94bcf52","ReBRAC-walker_walk-medium_expert-c1ac606c"]
0.5
false
-
false
-
-
-
-
-
-
2
0.2
Finished
adagrad
7h 54m 2s
-
0.1
0.0003
-
-
256
-
-
1000000
configs/sac-doup/walker_walk/expert.yaml
0.01
0.0003
true
-
expert
cuda
0.99
0.0003
-
-
1000
0
0.1
-
rebrac-walker_walk-expert-v2-sweep-v0
256
300000
-
-
-
10
["ReBRAC-walker_walk-expert-744643c6","ReBRAC-walker_walk-expert-a4f956b5","ReBRAC-walker_walk-expert-ad61f779","ReBRAC-walker_walk-expert-babde63c","ReBRAC-walker_walk-expert-d7482694"]
0.5
false
-
false
-
-
-
-
-
-
2
0.2
1-5
of 5
Cheetuh Run
Run set
8358
Humanoid Walk
Run set
8358
Add a comment
Created with ❤️ on Weights & Biases.
https://wandb.ai/tlab/ReBRAC/reports/ReBRAC-V-D4RL-Evaluation-for-Best-Hyperparameters--Vmlldzo0MzU5ODIx