Skip to main content
andrewzhang505
Projects
sample_factory
Reports
MuJoCo - Sample-Factory 2.0 vs CleanRL w/o EnvPool

MuJoCo - Sample-Factory 2.0 vs CleanRL w/o EnvPool

Andrew Zhang
Created on June 27|Last edited on July 27

2M4M6M8Mglobal_step020004000reward
: - Run set 3
Run set
10
Run set 3
5
Name
5 visualized
5
State
Notes
User
Tags
Created
Runtime
Sweep
anneal_lr
batch_size
capture_video
clip-vloss
clip_coef
clip_vloss
cuda
ent-coef
ent_coef
env_id
exp_name
gae
gae_lambda
gamma
learning-rate
learning_rate
max-grad-norm
max_grad_norm
minibatch_size
norm_adv
num-minibatches
num-steps
num_envs
num_minibatches
num_steps
seed
torch_deterministic
total_timesteps
track
update-epochs
update_epochs
vf-coef
vf_coef
wandb_project_name
charts/SPS
charts/avg_episodic_return
charts/episodic_length
charts/episodic_return
charts/learning_rate
global_step
losses/approx_kl
losses/clipfrac
losses/entropy
losses/explained_variance
Finished
-
costa-huang
3h 4m 7s
-
true
4096
false
-
0.2
false
true
-
0
Ant-v2
ppo_continuous_action
true
0.95
0.99
-
0.00295
-
3.5
1024
true
-
-
64
4
64
3
true
10000000
true
-
2
-
1.3
envpool-cleanrl
2774.6
-
1000
5271.17744
0.0000012085
9998336
0.0000012515
0
-4.44459
0.31732
1-1
of 1




Run set
10
Run set 3
5





Run set
10
Run set 3
5




Run set
10
Run set 3
5

For the environments below, CleanRL does not have benchmarks for non-EnvPool runs, so we are comparing against the sample efficiency of CleanRL with EnvPool

Run set
10
Run set 2
5



Run set
10
Run set 2
5



Run set
10
Run set 2
5



Run set
10
Run set 2
5



Run set
10
Run set 2
5


Created with ❤️ on Weights & Biases.

https://wandb.ai/andrewzhang505/sample_factory/reports/MuJoCo-Sample-Factory-vs-CleanRL-w-o-EnvPool--VmlldzoyMjMyMTQ0
Made with Weights & Biases. Sign up or log in to create reports like this one.