PG642 / multi-sample-factory

High throughput reinforcement learning on clusters
MIT License
2 stars 0 forks source link

Iss5 #7

Closed KonstantinRamthun closed 3 years ago

KonstantinRamthun commented 3 years ago

ONNX-export is not important enough, so we schedule it later. Added a script to use the enjoy function for the saving training scenario.

Closes #5