Ruan de Kock
Ruan de Kock
### What do you want to test? Unit test for the `MAPGEpochUpdateSeparateNetworks` and `MAPGMinibatchUpdateSeparateNetworks` components of the Jax PPO implementation that makes use of separate critic and policy netowrks. ###...
### What do you want to test? Unit test for the `MAPGWithTrustRegionStepSeparateNetworks` component of the Jax PPO implementation that makes use of separate critic and policy netowrks. ### Outline of...
### What do you want to test? Unit test for the `MAPGWithTrustRegionClippingLossSeparateNetworks` component of the Jax PPO implementation that makes use of separate critic and policy netowrks. ### Outline of...
## What? Updates to current Jax MAPPO implementation to match the Tensorflow MAPPO implementation. ## Why? We firstly want to be able to compare the Jax system against the Tensorflow...
## What? Full distributed Jax mappo system integration test. ## Why? Part of on going efforts to increase testing code coverage of callback redesigned Mava. ## Extra closes #542 In...
### Feature Have separate networks for the policy and critic in the Jax IPPO implementation. ### Proposal The current IPPO implementation has shared layers and separate heads for the policy...
### What do you want to investigate? Investigate/discuss whether multiple trainers should still be the default in Mava given the redesign or whether it should optional functionality. ### Definition of...
### Please describe what needs to be maintained? Investigate whether the use of enums is still relevant after Tensorflow support is deprecated. ### Describe the outcome you'd like When using...
The Reverse Training Starter Notebook runs fine up to 4000 episodes and then gives the following error: `Traceback (most recent call last): File "/usr/lib/python3.7/runpy.py", line 193, in _run_module_as_main "__main__", mod_spec)...
I am trying to run the the starter revere training notebook and run into the following issue: Everything seems to be working just fine, right up to when I want...