Skip to content

nhynes/abc

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

73 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Adversarial Behavioral Cloning

Improves on the SeqGAN idea by adding more reinfocement learning and GAN techniques like:

How to run

If you wish to enable Consensus Optimization (via the --grad-reg option), you'll need to patch PyTorch to allow forcing the use a twice-differentiable RNN.

python3 main.py will run the project with the default options. Output will be written to the run/ directory.

Shameless plug

The em tool makes it really easy to twiddle hyperparameters by tracking changes to code (no need to make everything an option!).

Just run em run -g 0 exp_name with your desired options and you'll find a reproducable snapshot in experiments/<exp_name>!

If you want to resume from a snapshot (perhaps with different options), use em resume -g 0 exp_name ...

You can also fork an experiment and its changes using em fork, but the quick and dirty solution is to run bash scripts/make_links.sh :)

Releases

No releases published

Packages

No packages published