|
NEML2 2.1.0
|
The most pervasive change in v2.1.2 is the removal of LabeledAxis. Variables no longer carry state/, forces/, residual/, or state/internal/ prefixes anywhere in input files.
Every option that previously accepted a prefixed variable name (e.g. tensor, variable, from, to, invariant, cauchy_stress, temperature, stress, etc.) now takes the bare variable name without any prefix.
"Old state" (history) variables, previously named under old_state/ or old_forces/ sub-axes, now use a ~N suffix where N is the lag order. For example:
| v2.1.1 name | v2.1.2 name |
|---|---|
| old_state/S | S~1 |
| old_forces/E | E~1 |
If you reference history variables explicitly (e.g. in a ComposedModel's additional_outputs list or a custom model), update those names accordingly.
In v2.1.1, NonlinearSystem inferred its unknowns from the model's state input sub-axis. In v2.1.2, you must list the unknown variables explicitly:
Residual variable names default to <unknown>_residual for each unknown. You can override them with the residuals option if needed:
These convenience driver types were removed. Use TransientDriver directly and supply the strain/stress controls through the generic force_SR2_* options.
Example: strain-control (formerly SDTSolidMechanicsDriver):
For mixed-control (formerly LDISolidMechanicsDriver), use the same pattern with the appropriate force_SR2_names/ic_SR2_names values for your control mode.
The predictor, custom_predictor, and custom_predictor_apply options have been removed from TransientDriver. Predictors are now declared as ordinary model blocks in [Models] and referenced directly from the ImplicitUpdate block via a new predictor option.
Before (v2.1.1):
After (v2.1.2):
| v2.1.1 driver option | v2.1.2 predictor type |
|---|---|
| predictor = 'PREVIOUS_STATE' | ConstantExtrapolationPredictor |
| predictor = 'LINEAR_EXTRAPOLATION' | LinearExtrapolationPredictor |
| custom_predictor = '<name>' | keep the existing model block; wire it with predictor = '<name>' on ImplicitUpdate |
| (no predictor in driver) | ConstantExtrapolationPredictor |