Hi all, any news on that topic?
I’m creating the trainer in separate custom process (on the same node where ray had been initialized) and it also gets stuck on compute_action().
I’m running 1.11.0 and just creating a local worker. Anything I could try?
Thanks