My Ray programs stops learning when using distributed compute

[Stefan-1313]

For anyone ending up here, please checkout other training architectures. SimpleQTrainer is just too simple and does not support this, but others do!

@Stefan-1313

I cannot find such a parameter, but I look for a parameter like num_train_batches_after_sample_collection. This parameter simply controls the number of policy updates when a policy update is performed.

The simplest one that supports the above is: DQNTrainer. Try this one instead of SimpleQTrainer. The parameter to control the train/experience-collection ratio is config_simple["training_intensity"].

Good luck!