danijar/dreamer issues and pull requests

#58 - model training fails

Issue - State: open - Opened by sumwailiu 20 days ago

#57 - Different std of models

Issue - State: open - Opened by elisaparga19 10 months ago

#57 - Different std of models

Issue - State: open - Opened by elisaparga19 10 months ago

#56 - KL clipping: before or after averaging?

Issue - State: closed - Opened by vmoens over 2 years ago - 1 comment

#56 - KL clipping: before or after averaging?

Issue - State: closed - Opened by vmoens over 2 years ago - 1 comment

#55 - Provided scores don't match the results

Issue - State: closed - Opened by zhixuan-lin over 2 years ago - 2 comments

#55 - Provided scores don't match the results

Issue - State: closed - Opened by zhixuan-lin over 2 years ago - 2 comments

#54 - Can't reproduce results in some environments

Issue - State: closed - Opened by zplizzi almost 3 years ago - 3 comments

#53 - freenats inconsistent with tf1 repo

Issue - State: closed - Opened by ssnl about 3 years ago - 1 comment

#52 - mbrl distractors

Pull Request - State: open - Opened by homangab over 3 years ago

#51 - the code is running without any results and output

Issue - State: closed - Opened by lylyjy over 3 years ago - 5 comments

#50 - AttributeError: 'MirroredStrategy' object has no attribute 'experimental_run_v2'

Issue - State: closed - Opened by lylyjy over 3 years ago - 1 comment

#49 - my.hackmit.org Can't register

Issue - State: closed - Opened by Bobolovewill over 3 years ago - 1 comment

#48 - slow in atari tasks

Issue - State: closed - Opened by xiangyyyy almost 4 years ago - 2 comments

#47 - How to run on short episodes?

Issue - State: closed - Opened by mctigger almost 4 years ago - 2 comments

#46 - What is this line for?

Issue - State: closed - Opened by mctigger almost 4 years ago - 1 comment

#45 - Differences in free nats clipping between Dreamer, early and final PlaNet implementation

Issue - State: closed - Opened by TyPh00nCdr almost 4 years ago - 2 comments

#44 - Free nats over batch and time dimension?

Issue - State: closed - Opened by mctigger almost 4 years ago - 1 comment

#43 - Runtime performance

Issue - State: closed - Opened by mctigger almost 4 years ago - 1 comment

#42 - Spikes in Loss?

Issue - State: closed - Opened by mctigger almost 4 years ago - 2 comments

#41 - difference between "CheetahRun-v0" on DM vs "half-cheetah-v2" on Mujuco

Issue - State: closed - Opened by randyhouse almost 4 years ago - 1 comment

#40 - lost of file 'dm_control'

Issue - State: closed - Opened by yueyue-yue about 4 years ago - 3 comments

#39 - Invalid one-hot action with Google Research football environment

Issue - State: closed - Opened by jgonik about 4 years ago - 1 comment

#38 - Tensorflow-probability version

Issue - State: closed - Opened by eugenevinitsky about 4 years ago - 2 comments

#37 - Fix wrappers bug

Pull Request - State: open - Opened by mohith-sakthivel over 4 years ago

#36 - A question about reward and observation pairing in wrapper

Issue - State: closed - Opened by FinnWeng over 4 years ago - 2 comments

#35 - Performance on sparse reward

Issue - State: closed - Opened by yiwc over 4 years ago - 2 comments

#34 - Dreamer for DMLab

Issue - State: closed - Opened by syigzaw over 4 years ago - 1 comment

#33 - Dreamer for Atari

Issue - State: closed - Opened by michaelzhiluo over 4 years ago - 4 comments

#32 - Sudden performance drop when value_grad_norm returns Inf

Issue - State: closed - Opened by tung-nd over 4 years ago - 8 comments

#31 - Is there a JavaScript or typescript version out there?

Issue - State: closed - Opened by NullVoxPopuli over 4 years ago - 1 comment

#30 - Add flag for slow target value function

Issue - State: closed - Opened by AliengirlLiv over 4 years ago - 1 comment

#29 - How to run Atari environments?

Issue - State: closed - Opened by Jannkar over 4 years ago - 2 comments

#28 - Minor bug fixes

Pull Request - State: open - Opened by markelsanz14 over 4 years ago

#27 - Do not override environment variables if they are explicitly set

Pull Request - State: open - Opened by asolano over 4 years ago

#26 - Process parallel tf scalar summary hangs

Issue - State: closed - Opened by AlexanderKoch-Koch over 4 years ago - 4 comments

#25 - Upgrade tensorflow-gpu to 2.2.0

Pull Request - State: closed - Opened by letusfly85 over 4 years ago - 2 comments

#24 - Why does dreamer not require stored hidden states?

Issue - State: closed - Opened by xlnwel over 4 years ago - 7 comments

#23 - Running Atari games seems to crash

Issue - State: closed - Opened by snailrowen1337 over 4 years ago - 2 comments

#22 - Is RSSM trained by forward prediction or reconstruction?

Issue - State: closed - Opened by zafarali over 4 years ago - 1 comment

#21 - Why not use bit-depth preprocessing?

Issue - State: closed - Opened by cross32768 over 4 years ago - 1 comment

#20 - Have you meet value function explosion when developing?

Issue - State: closed - Opened by IcarusWizard over 4 years ago - 1 comment

#19 - Memory leak when calling policy function?

Issue - State: closed - Opened by AntoineRichard over 4 years ago - 3 comments

#18 - Why actor-critic loss is weighted by accumulated discount?

Issue - State: closed - Opened by IcarusWizard over 4 years ago - 1 comment

#17 - pcont when running Atari Games

Issue - State: closed - Opened by Mohan-Zhang-u over 4 years ago - 1 comment

#16 - How to train on low-dimensional Gym envs?

Issue - State: closed - Opened by jarlva almost 5 years ago - 1 comment

#15 - How to run the code without a GPU?

Issue - State: closed - Opened by jarlva almost 5 years ago - 2 comments

#14 - Missing Python dependencies

Issue - State: closed - Opened by filangelos almost 5 years ago - 5 comments

#13 - Isolating environments into threads or processes

Issue - State: closed - Opened by Mehooz almost 5 years ago - 1 comment

#12 - How to tune the hyper parameters of new RL algorithms?

Issue - State: closed - Opened by xlnwel almost 5 years ago - 6 comments

#11 - Relaxed categorical for gradients through discrete actions

Issue - State: closed - Opened by xlnwel almost 5 years ago - 1 comment

#10 - Why are the losses sometimes reported to be infinity?

Issue - State: closed - Opened by yuqingd almost 5 years ago - 2 comments

#9 - Running Dreamer on Gym environments

Issue - State: closed - Opened by arunavo4 almost 5 years ago - 1 comment

#8 - Is there a shift in the action sequence?

Issue - State: closed - Opened by IcarusWizard almost 5 years ago - 2 comments

#7 - Is distributed training supported?

Issue - State: closed - Opened by arunavo4 almost 5 years ago - 12 comments

#6 - Why are the random sampling ops patched?

Issue - State: closed - Opened by danijar almost 5 years ago - 1 comment

#5 - Why does the RSSM receive different inputs for learning and imagination?

Issue - State: closed - Opened by danijar almost 5 years ago - 3 comments

#4 - Why is the action mean bounded before applying tanh?

Issue - State: closed - Opened by xlnwel almost 5 years ago - 4 comments

#3 - Are policy and value gradients propagated back through the world model?

Issue - State: closed - Opened by xlnwel almost 5 years ago - 2 comments

#2 - What is the meaning of pcont?

Issue - State: closed - Opened by xlnwel almost 5 years ago - 9 comments

#1 - Could you comment and format the code?

Issue - State: closed - Opened by FrancescoSaverioZuppichini almost 5 years ago - 1 comment

GitHub / danijar/dreamer issues and pull requests