goku's Introduction

GOKU - Deep Generative ODE Modelling with Known Unknowns

This repository is an implementation of the GOKU paper: Generative ODE Modeling with Known Unknowns.

Data creation

To create the datasets used in the paper run:

Friction-less pendulum: python3 create_data.py --model pendulum
Friction pendulum: python3 create_data.py --model pendulum --friction
Double-pendulum experiment: python3 create_data.py --model double_pendulum
Cardiovascular system: python3 create_data.py --model cvs

The data would be created using default arguments. To view / modify them check the file config.py, and create_data.py.

Training

To train the GOKU model run: python3 goku_train.py --model <pendulum/pendulum_friction/double_pendulum/cvs>

To train baselines:

Latent-ODE: python3 latent_ode_train.py --model <pendulum/pendulum_friction/double_pendulum/cvs>.
LSTM: python3 lstm_train.py --model <pendulum/pendulum_friction/double_pendulum/cvs>.
Direct-Identification (DI) has 3 different files for the different datasets (it cannot run the friction pendulum, since it needs the entire ODE functional form):
- Pendulum: python3 di_baseline_pendulum.py
- Double Pendulum: python3 di_baseline_double_pendulum.py
- CVS: python3 di_baseline_cvs.py

Requirements:

python 3
pytorch
numpy
gym (for the pendulum and double pendulum experiments)

goku's People

Contributors

Stargazers

Watchers

goku's Issues

Error generating pendulum data

gym version 0.11.0
scikit-image version 0.16.2

Is this error due to scikit image/gym or an error in the preproc() function code? Any pointers towards solving this would help and/or sharing the pendulum data with a drive link would help too! Thanks.

python3 create_data.py --model pendulum
  0%|                                                                     | 0/500 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "create_data.py", line 130, in <module>
    make_dataset(args)
  File "create_data.py", line 49, in make_dataset
    raw_data, latent_data, params_data = args.create_raw_data(args)
  File "/home/pranav/GIT_repos/GOKU/utils/create_pendulum_data.py", line 78, in create_pendulum_data
    processed_frame = preproc(env.render('rgb_array'), side)
  File "/home/pranav/GIT_repos/GOKU/utils/create_pendulum_data.py", line 19, in preproc
    return skimage.transform.resize(X, [int(side), side]) / 255.
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/skimage/transform/_warps.py", line 166, in resize
    preserve_range=preserve_range)
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/skimage/transform/_warps.py", line 807, in warp
    raise ValueError("Cannot warp empty image with dimensions", image.shape)
ValueError: ('Cannot warp empty image with dimensions', (0, 0))
Exception ignored in: <bound method Viewer.__del__ of <gym.envs.classic_control.rendering.Viewer object at 0x7fcf3301eda0>>
Traceback (most recent call last):
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/gym/envs/classic_control/rendering.py", line 143, in __del__
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/gym/envs/classic_control/rendering.py", line 62, in close
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/pyglet/window/xlib/__init__.py", line 480, in close
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/pyglet/gl/xlib.py", line 345, in destroy
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/pyglet/gl/base.py", line 334, in destroy
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/pyglet/gl/xlib.py", line 335, in detach
  File "/home/pranav/anaconda3/envs/my_lab/lib/python3.6/site-packages/pyglet/gl/lib.py", line 97, in errcheck
ImportError: sys.meta_path is None, Python is likely shutting down

Recommend Projects