Is your feature request related to a problem? Please describe.
For diffusion downscaling we need a sampling strategy, where sampling should:
- traverse the high-noise regime quickly;
- spend most computation in the low-noise regime where fine-scale fidelity is determined.
Describe the solution you'd like
Introduce a piecewise sigma schedule:
- rapid collapse through the high-noise regime;
- Karras-style refinement in the low-noise regime.
This is intended as a purely additive change consisting of one new sampling schedule in models/src/anemoi/models/transport/schedules.py
Describe alternatives you've considered
No response
Additional context
No response
Organisation
No response
Is your feature request related to a problem? Please describe.
For diffusion downscaling we need a sampling strategy, where sampling should:
Describe the solution you'd like
Introduce a piecewise sigma schedule:
This is intended as a purely additive change consisting of one new sampling schedule in models/src/anemoi/models/transport/schedules.py
Describe alternatives you've considered
No response
Additional context
No response
Organisation
No response