huanglangwen

Need new ways to specify parameters for adjoint sensitivity analysis

10

I think it is better to pass in `f_p`, `g_p` and `g_u` explicitly for adjoint sensitivity analysis instead of relying on the parameter `p` in `f(du,u,p,t)` and `dg(out,u,p,t)`. Because `p`...

Handling of missing values

4

When loading grib files with values around 9999, those values are masked to np.nan by cfgrib in the following line: https://github.com/ecmwf/cfgrib/blob/0834b19d56ab2a817cc5a3309149806afdcf0ced/cfgrib/dataset.py#L358 Is there a way to set missing values manually...

Unexpected results when having symbolic shaped arrays and negative indices

8

**Describe the bug** Got unexpected results when having symbolic shaped arrays and negative indices. **To Reproduce** ```python import numpy as np import dace M = dace.symbol('M') N = dace.symbol('N') @dace.program...

oop jacobian can't preserve array-like objects

5

`forwarddiff_color_jacobian` automatically converts array-like input object into array. When `f` depend on specific property of the input object, the jacobian will fail. MWE: ```julia using OrdinaryDiffEq, SparseDiffTools mutable struct SimTypeg{T,T2}

Enable preconditioning

1

Preconditioned Newton-Krylov solver seems provide better performance than Exponential-Krylov integrator without preconditioning. It might be a good idea to add preconditioning to `expv` using the SAI method. [1] https://ris.utwente.nl/ws/portalfiles/portal/5092923/memo1992.pdf [2]...