Releases: DiffEqML/torchdyn
Releases · DiffEqML/torchdyn
API redesign, more models, numerical methods!
The current release greatly extends the functionality offered by torchdyn
, planning for a larger 1.0
release later this month (with multiple shooting layers, root finding, rudimentary implicit solvers, hybrid neural DE solvers).
- New
ODEProblem
for standard IVPs. Alternative sensitivity algorithms relyODEProblems
for sanity, and are specified at__init__
. Backpropagating through the functional API (direct calls toodeint
) only uses reverse-mode autodiff. - A new interpolated adjoint variant relying on
torchcde
utilities. - New functional
odeint
API with adaptive-step solvers:dopri5
,tsit5
. - New
odeint_hybrid
for multi-mode hybrid ODEs with stochastic or deterministic events. - Seamless hypersolver integration with
odeint
API, with same call signature!
Some more minor or less interesting changes:
- Streamlining of tests and new tutorial automated notebook validation.
s
changed tot
for more clarity and compatibility with the current literature.
Expect bugs and rough edges (and some WIP hidden in the library). The 1.0
will clean those up.
torchdyn goes stochastic
v0.2.2.1 [org] update publish
torchdyn prepares for SDEs
v0.2.1 [tests] adjust tolerance of adjoint tests
torchdyn goes generative
v0.2.0 Update change_log.rst