agents

agents copied to clipboard

PyEnvironment Methods Incompatible with TF

7

The docstring for `tf_py_environment.__getattr__` indicates that certain PyEnvironment methods might be incompatible with TF. ```python def __getattr__(self, name: Text) -> Any: """Enables access attributes of the wrapped PyEnvironment. Use with...

ergsfe

Feature Request: support Priority Replay buffer with reverb in a parallel environment

8

First of All, Thank you for this awesome repo, It saved months of my life ;-) I'm just letting you know that I'm following closely the ongoing implementation of the...

Axel-CH

Extracting additional variables from the environment?

1

Hi, I have been adapting the DQN tutorial file for a.custom environment. It seems to learn fine, however I have an additional metric that I want to extract and plot...

JohnBurden

Add Adavanced actor critic agent

1

Several Deep RL agents that are missing such as A2C, A3C which can be added. Further work could also adding MARL agents such as MAA2C or MADDPG

veds12

Feature Request: Support Duel DQN

1

A feature request proposal to add support of Duel DQN, as suggested in [paper](https://arxiv.org/pdf/1511.06581.pdf) [Dueling Network Architectures for Deep Reinforcement Learning] , which is described as: " The main benefit...

king821221

Tutorial 1: Train a Deep Q Network with TF-Agents and the use of a driver

1

Hello, I believe the tutorial 1_dqn_tutorial.ipynb has an unnecessary import of dynamic_step_driver. The module is not used at all, so the scripts runs just fine when commented out. Furthermore, I...

jan-gebauer

Error in loading policy from saved model.

9

I was trying to load a saved model policy using the following code, but it ends with errors pasted below `saved_policy = tf.saved_model.load('policy_0') ` Software versions: - Tensorflow =2.3.0 -...

n3011

Does CheckPointer load previous agent, policy, and replay buffer data from saved files?

6

Hello, There is a tutorial on `Checkpointer`. https://github.com/tensorflow/agents/blob/master/docs/tutorials/10_checkpointer_policysaver_tutorial.ipynb In `Checkpointer` and `Restore checkpoint` sections, I found the following code. ``` train_checkpointer = common.Checkpointer( ckpt_dir=checkpoint_dir, max_to_keep=1, agent=agent, policy=agent.policy, replay_buffer=replay_buffer, global_step=global_step )...

cosmir17

Should time_step_spec of an array valued reward return an array valued discount?

1

The time_step_spec function only takes observation_spec and reward_spec array specifications, but if the reward_spec specifies a multidimensional array, shouldn't the discount_spec match its shape or at least accept an argument...

guachoperez

Function to Make Predictions From Policy

6

Suppose I save some policy: ` saver = PolicySaver(my_policy, batch_size=None) ` ` saver.save('my_policy') ` And then I load this policy ` policy = tf.saved_model.load('../DATA/Multi/policy_9') ` Now I _**ONLY**_ want the...

lukepolson