Skip to main content
costa-huang
Projects
cleanRL
Reports
descriptiveness

descriptiveness

Costa
Created on July 16|Last edited on August 10
study batch size

5001k1.5k2kStep0246810
5001k1.5k2kStep0.10.20.30.4
5001k1.5k2kStep0123
Select runs that logged ppo/objective/score_total
to visualize data in this line chart.
5001k1.5k2kStep0.0010.0020.0030.004
5001k1.5k2kStep00.0020.0040.0060.008
batch_size=64
1
batch_size=128
1
batch_size=256
1
batch_size=512
1
batch_size=512, adam 1e-5
1
oai
10
batch_size=512, adam 3e-5
1
oai 2
batch_size=512, adam 8e-5
1
batch_size=512 sgd
1
batch_size=512, adam 1e-4
1
batch_size=512, 2e-4
1
batch_size=512 init_kl_coef=0.2
1
batch_size=512 init_kl_coef=0.25
2
batch_size=512, adam 8e-5, init_kl_coef=0.25
1
batch_size=512, adam 4e-5, init_kl_coef=0.3
1
Run set 17
1
bigger KL penalty
1
Run set 19
4
Run set 20
10
Run set 21
18
new adam
8
tensorflow-style adam
10
tensorflow-style adam 2
10
tensorflow-style adam gpt2-medium
10
gpt2-large
1
gpt2
10
gpt2-xl
10
gpt2-xl PT
10
gpt2-medium
20
gpt2-large
20
Run set 31
0
gpt2 PT
10
Run set 34
31
gpt2 fix clip val
10
alt clip val fix
10
Run set 37
1
gpt2 fix clip val
10
Run set 39
10
Run set 40
10



Created with ❤️ on Weights & Biases.

https://wandb.ai/costa-huang/cleanRL/reports/descriptiveness--Vmlldzo0ODkzNjE2
Made with Weights & Biases. Sign up or log in to create reports like this one.