Anoop Korattikara

Hi, This docstring is just meant to indicate that the methods in PyEnvironment do not accept/return Tensors or do other things that are not TF compatible. But calling .seed on...

@summer-yue Can you PTAL at this?

Yes one option is to extend the TimeStep class, e.g. you can add an info field. Or you can add a property in the environment that stores the last value...

Can you try these: https://github.com/tensorflow/agents/blob/master/tf_agents/policies/policy_loader.py https://github.com/tensorflow/agents/blob/bdf6694dec482f8f894ee61dea750e1c0d48ad3e/tf_agents/policies/py_tf_eager_policy.py#L102

Hi Sean, Trying to understand your question more. Is your code significantly different from our DQN example: https://github.com/tensorflow/agents/blob/master/tf_agents/agents/dqn/examples/v2/train_eval.py This shows how to create the global step and saves in the...

Hi Sean, Yes the train_checkpointer.initialize_or_restore() in the dqn train_eval I linked is meant to restore the agent, policy variables etc. https://github.com/tensorflow/agents/blob/5ec6898513ed31cdac6819604536a410e1a948d3/tf_agents/agents/dqn/examples/v2/train_eval.py#L220 It can be killed, and when restored, it should...