torchdyn
torchdyn copied to clipboard
DiffEqML
A PyTorch library entirely dedicated to neural differential equations, implicit models and related numerical methods
ALF ([AsynchronousLeapfrog](https://github.com/DiffEqML/torchdyn/blob/685e4483cec7b884fd3d900208913140c35d86c0/torchdyn/numerics/solvers.py#L96)) is currently unfinished. With ALF, [MALI ](https://arxiv.org/abs/2102.04668) can be implemented and benchmarked. [Here ](https://github.com/juntang-zhuang/TorchDiffEqPack/tree/master/TorchDiffEqPack/odesolver) is a good starting point. @juntang-zhuang in case there are issues with this.
Good first issue for someone interested in getting familiar with diffusion models and [score-matching SDEs](https://openreview.net/forum?id=PxTIG12RRHS). Those interested are encouraged to get in touch directly as we have internal private prototypes...
Extended tutorial on using [MSLs](https://arxiv.org/abs/2106.03885) to speed up Neural CDEs via root tracking.
We run an automated script in `test/validate_tutorials.py` to ensure all notebooks are always runnable. Tutorials in modules not included in the script might not run due to API changes. Older...
We plan to kickstart our pretrained model zoo with weights of a SOTA CNF trained on CIFAR10 and ImageNet.
Good first issue for someone interested in getting familiar with implicit models and DEQs. Those interested are encouraged to get in touch directly as we have internal private prototypes that...
When looking at the source code under /torchdyn/core/neuralde.py, I found that the NeuralODE class is missing defunc property. And this bug results in an error for /tutorials/module4-model/m4e_lagrangian_nets.ipynb. The notebook can...
Opening the first PR for `NeuralSDE` for integrating sdeint into torchdyn. Also, as I formatted code files I have made changes using [black](https://github.com/psf/black), we can use this format for future...
Hi, I noticed that my code fails when using two instead of only one optimizer for different groups of parameters. The minimal example below is from your repo, but I...
