Analysis and Sampling of Molecular Simulations by adversarial Autoencoders
Binder is likely to be broken right now :-(
TODO
TODO
TODO
Full support for distributed hyperparameter tuning is available at CERIT-SC Jupyterhub:
- Go to https://hub.cloud.e-infra.cz/ and log in with your Metacentrum account
- Either click on Start my server for the default, or type a specific name and click Add New Server
- Fill in the submit form:
- Select an image: Custom
- Custom image name: ljocha/asmsa:2023-19
- Select persistent home type: New at the first time, Existing is prefered afterwards
- Select persistent home (when Existing in the previous choice): pick you prefered one, or stick with the only one offered
- I want MetaCentrum home: no
- Would you like to connect project directory: no
- Select number of CPUs: 2 (it's enough for the notebook, parallel workers are not counted here)
- Memory: 8 GB (appears to be enough for this usecase)
- GPU: None (it turns out that our models are too small to leverage GPU accelleration)
- Click on Start
- Depending on the container image cache status Jupyterlab starts in few seconds (the image was cached) or several minutes (it must be downloaded)
- click on asmsa.ipynb in the left panel and follow instructions in the notebook
TODO (screenshots, ...)
TODO
TODO
TODO
TODO
In development... No main usable version released yet