Kaixhin / PlaNet

Deep Planning Network: Control from pixels by latent planning with learned dynamics

Kaixhin/PlaNet Issues

Hyperparameter ballpark for symbolic envs
Closed 5 years ago3
Potential error in the latent overshooting objective
Closed 3 years ago2
Terminal State handling
Closed 4 years ago1
Difference in decoder model from original paper?
Closed 4 years ago4
Are RNN hidden states ever reset?
Closed 4 years ago2
Action clamp not being applied
Closed 4 years ago
Question about the overshooing KL divergence cost scaling.
Closed 4 years ago1
Clipping of actions
Closed 4 years ago5
Need testing?
Closed 5 years ago26
how long does it take for each environment to train?
Closed 4 years ago1
Write custom serialisation for experience replay
Updated 4 years ago9
Hyper-parameter
Closed 5 years ago8
Test
Closed 5 years ago
Questions
Closed 5 years ago1
allow other embedding-size besides 1024
Closed 5 years ago5
Random Chunk Sample Vs Random idx sample
Closed 5 years ago1
pyglet error (No display)
Closed 5 years ago3
Reconstruction video is glitchy
Closed 5 years ago7
GPU program failed to execute at ___
Closed 5 years ago2
Altering Core Components
Closed 5 years ago1
Parallelise testing
Closed 5 years ago
Add test phase
Closed 5 years ago
Save and reload experience replay
Closed 5 years ago
Nonlinearity flag
Closed 5 years ago
Vectorise code
Closed 5 years ago
Implement DELIP
Updated 5 years ago