Repository Issue Activity (beta)

#1304 Learning rate does not decay in examples/vizdoom_ppo.py when using ICM wrapper

#1057 Code duplication between ReplayBuffer and ReplayBufferManager

#1302 New feature request: reward-hacking onset monitoring hooks

#1056 Improve and extend Documentation Content

#1301 TD3Params `policy_noise` docstring

#1215 Towards Roadmap

#1034 Clearer separation between the trainer and the algorithm and refactoring of policy classes

#1053 Update Gymnasium to v1.0.0a1

#1036 Rename state_shape to obs_shape

#1299 Reconsider closed issues for Tianshou v3

#1297 Loading policy from persitence dir

#1291 Collector.collect() crashes with negative collect_time due to time.time() clock skew

#1289 Enabling extra eval produces error due to unconstrained pandas version

#1288 `ExperimentResult.trainer_result` appears to always be `None` — possible missing assignment in `Experiment.run()`?

#1271 Implement GRPO

#1222 Wandb logger

#1286 I have idea for predict the future events

#983 gymnasium.error.NameNotFound: Environment `PongNoFrameskip` doesn't exist.

#1285 The examples under examples/atari in version 2.0.0 run very slowly, and the test reward does not change with training.

#895 Need a flexible method to record training data in TensorBoard

#1282 No module tianshou.algorithm

#1281 Buffer observation equals to next observation using ShmemVectorEnv

#1269 link to contributing docs is broken

#1274 Incompatibility with Python 3.13 due to outdated packages requirement

#1198 Handling LR scheduling when episodes instead of steps are collected