physicsnemo
physicsnemo copied to clipboard
⛰️[EPIC]: CorrDiff Support & Enhancements
### Code structuring
- [ ] Split generate.py, train.py and training_loop.py into functions/classes that handle more well-defined tasks
- [ ] Reduce redundancy between the above files
- [x] Clean up handling of config files (https://github.com/NVIDIA/modulus/pull/355)
- [x] Specify clearly in an interface what interface data loaders need to follow, to facilitate adding new loaders (https://github.com/NVIDIA/modulus/pull/376)
- [ ] Same as above, but for models (e.g. U-Nets)
- [ ] Interoperability with diagnostic model code base (TBA)
- [x] Dump checkpoints, logs, and code in separate random directories per run (https://github.com/NVIDIA/modulus/pull/374)
### Optimizations
- [ ] https://github.com/NVIDIA/modulus/issues/424
- [ ] Support distributing generation over time steps as well as ensemble members
### Features
- [ ] Enable per-channel loss
- [ ] https://github.com/NVIDIA/modulus/issues/394
- [ ] https://github.com/NVIDIA/modulus/issues/416
- [ ] https://github.com/NVIDIA/modulus/issues/392
- [ ] Compute loss in inference
- [ ] https://github.com/NVIDIA/modulus/issues/425
- [ ] https://github.com/NVIDIA/modulus/issues/440
### QA
- [x] Make a subset of the CWA dataset
### Misc
- [ ] https://github.com/NVIDIA/modulus/issues/367
- [ ] https://github.com/NVIDIA/modulus/issues/393
### Go-to-market readiness
- [ ] https://github.com/NVIDIA/modulus/issues/407
- [ ] https://github.com/NVIDIA/modulus/issues/408
- [ ] https://github.com/NVIDIA/modulus/issues/409
### Improvements
- [ ] https://github.com/NVIDIA/modulus/issues/414
- [ ] https://github.com/NVIDIA/modulus/issues/415
- [ ] https://github.com/NVIDIA/modulus/issues/417
- [ ] https://github.com/NVIDIA/modulus/issues/441
### Bugs
- [ ] https://github.com/NVIDIA/modulus/issues/418
### Issues
- [ ] https://github.com/NVIDIA/modulus/issues/437
#376 is merged so we can probably check off the box "Specify clearly in an interface what interface data loaders need to follow, to facilitate adding new loaders".
The remaining items will be tracked in this Epic: https://github.com/NVIDIA/modulus/issues/589