Search code examples
pythonreinforcement-learningtensorboard

Tensorboard stops updating in Google Colab during learning with stable baselines


I am using PPO stable baselines in Google Colab with Tensorboard activated to track the training progress but after around 100-200K timesteps tensorboard stops updating even with the model still training (learning), does anyone else have this issue and know a fix for it?


Solution

  • stable baselines doesnt seem to run well on CoLab because of the need to downgrade to tensorflow 1.6 which doesnt run well with tensorboard so instead I used to the newer stable baselines3 with current tensorflow version and tensorboard works fine.