Xingdong Zuo

icon indicating a website link https://zuoxingdong.github.io/

icon company NAVER Corp icon location Seongnam, South Korea icon location AI in well-being is my dream. Neural networks need to understand the world causally.

Results 22 issues of Xingdong Zuo

Logging: convert loaded loggings entirely into pandas `DataFrame`

Load the entire logging folder including multiple configurations and multiple random runs. All the post-processings are performed on the DataFrame, e.g. smoothing the episode returns. e.g. ID | lr |...

design
refactor

Refactor Hyperparameter classes

Inspired by Amazon SageMaker and Ray, try to refactor the classes for hyperparameters. e.g. - Categorical - Continuous - Logarithmic: for small scale, e.g. learning rate

design
refactor

Support Python 3.8

- In the script folder: remove the "typing" package, it is not necessary anymore.

enhancement

Migrate obs/reward normalization from env.wrappers into Agent itself

- Make online statistics as `nn.Parameter` and registered inside the module. It becomes trackable - Similar style with how the BatchNorm is implemented in PyTorch - Different behavior between train/eval...

design
refactor

Integrate Ray as an additional backend for parallel experiments

As an additional backend to Python Multiprocessing.

design

Replace from_numpy/tensorify with as_tensor, remove tensorify

Merge seed and device into config object

design

Refactor Env: use gym.Env

Support metadata: `reward_range`, `spec` etc

Add Sokoban environment

todo

Fix Box space: recast to np.float32 to avoid warning in gym 0.15.6

[These two lines](https://github.com/zuoxingdong/dm2gym/blob/ad4e4edb975ba2ffb766d663ff77b8a1c6c5f3d6/dm2gym/envs/dm_suite_env.py#L11-L12)