* Proper dimensions for entropy, sum before bonus in PPO
* Make entropy reporting same as TF
* Always use separate critic
* Revert to shared
* Remove unneeded extra line
* Change entropy shape in test
* Change another entropy shape
* Add entropy summing to evaluate_actions
* Add notes about torch.abs(policy_loss)
* Torch setup.py
* Set torch to default
* Make torch default in setup.py
* Remove indents
* Remove other instances of TF being used
* Add tensorboard to setup.py
* Adding correst setup commands for verifying torch is installed (#4524)
* Adding correst setup commands for verifying torch is installed
* Editing the test_requirments to add tf and remove torch
* Develop torchdefault raise outside setup (#4530)
* Torch not imported error to raise at first usage
* Torch not imported error to raise at first usage
* [refactor] Use PyTorch TensorBoard utils (#4518)
* Convert stats writer to use PyTorch TB support
* Use common function to print params
* Update test
* Bump tensorboard to 1.15 to fix the tests
* putting tensorboard 1.15.0 as min version requirement
Co-authored-by: vincentpierre <vincentpierre@unity3d.com>
* [Docs] Initial documentation changes for making...
* Updates the ML-Agents Docker image based on things
we have learned from training in the cloud.
* Automatically pushes the image to our own GCP projects
on each release.
* Move folders around
* [Make tests to compare compressed and non-compressed obs
> Make `Utilities.TextureToTensorProxy` public (debatable)
> GridSensor can now be compressed or uncompressed
> Added 2 scenes and an integration test to make sure compressed and uncompressed obs are the same
* fix typo
* renaming file so pytest will not try to run it
* rename yamato config file
* Need standalone build for 2019.4
* Running only on 2018.4 since it seems to be the only version thhat can build
* Typo in the name of the testing script
* Renaming the scene
So this is funny Python bug but
`a.strip(".unity")` will remove the last letter of `a` (before the .unity) if the last letter is "t"
not kidding, try :
```
a = "fewgfwegwrgvrt.unity"
a.strip(".unity")
```
* using splitext rather than strip to find executable name`
* Rename and move TextureToTensor
* Addressing comments
* re...
* Destroy stepper when playmode change
* Detroy stepper if it does not belong to the current Academy
Co-authored-by: Chris Elion <chris.elion@unity3d.com>
VisualFoodCollector is now an example environment of using a mix of visual and vector observation and is able to train with default config file.
Co-authored-by: Vincent-Pierre BERGES <vincentpierre@unity3d.com>
Added stacking to multi-dimensional and compressed observations and added compressed channel mapping in communicator to support decompression.
Co-authored-by: Vincent-Pierre BERGES <vincentpierre@unity3d.com>
Co-authored-by: Chris Elion <chris.elion@unity3d.com>
* Moved components to the tf folder and moved the TrainerFactory to the `trainer` folder
* Addressing comments
* Editing the migrating doc
* fixing test
* Fixing CPU resource allocation for new CI
* Update ml-agents/mlagents/torch_utils/cpu_utils.py
Co-authored-by: Ervin T. <ervin@unity3d.com>
* [skip ci] testing
* use shares only if on kubernetes
Co-authored-by: Ervin T. <ervin@unity3d.com>
* initial commit
* works with Pyramids
* added unit tests and a separate config file
* Adding first batch of documentation
* adding in the docs that rnd is only for PyTorch
* adding newline at the end of the config files
* adding some docs
* Code comments
* no normalization of the reward
* Fixing the tests
* [skip ci]
* [skip ci] Make sure RND will only work for Torch by editing the config file
* [skip ci] Additional information in the Documentation
* Remove the _has_updated_once flag
* Don't run value during inference
* Execute critic with LSTM
* Address comments
* Unformat
* Optimized soft update
* Move soft update to model utils
* Add test for soft update
GitHub
Ruo-Ping Dong
Ervin Teng