Feature: Converge Thresholds #699
Conversation
There was a problem hiding this comment.
Hi @AndresOrtegaGuerrero , thanks for the work! I made a few small suggestions.
- The
stepofscf_conv_thretc., is better to be similar to the default value.
I already approved the PR. Please make the changes and merge the PR. There is no need for me to review it again.
| num_atoms = 1 | ||
| if self.input_structure is not None: | ||
| ase_structure = self.input_structure.get_ase() | ||
| num_atoms = ase_structure.get_global_number_of_atoms() |
There was a problem hiding this comment.
| num_atoms = 1 | |
| if self.input_structure is not None: | |
| ase_structure = self.input_structure.get_ase() | |
| num_atoms = ase_structure.get_global_number_of_atoms() | |
| num_atoms = len(self.input_structure.sites) if self.input_structure else 1 |
| @@ -357,7 +441,7 @@ def set_metallic_magnetization(self, parameters): | |||
|
|
|||
| def set_panel_value(self, parameters): | |||
| """Set the panel value from the given parameters.""" | |||
|
|
|||
| # protocol_parameters = PwBaseWorkChain.get_protocol_inputs(protocol) | |||
| self.scf_conv_thr = ipw.BoundedFloatText( | ||
| min=1e-15, | ||
| max=1.0, | ||
| step=0.000001, |
There was a problem hiding this comment.
| step=0.000001, | |
| step=1e-10,, |
| self.forc_conv_thr = ipw.BoundedFloatText( | ||
| min=1e-15, | ||
| max=1.0, | ||
| step=0.00001, |
There was a problem hiding this comment.
| step=0.00001, | |
| step=0.0001, |
|
The user may want to change the value by mouse; in the case of SCF conv., when I click the up arrow, I assume the user wants to increase from 8e-10 to 9e-10, but it increases too much. As shown here: qeapp-conv.mp4 |
for more information, see https://pre-commit.ci
I did the changes you suggested and it helps on this, however the issue is that the default values depend on the structure selected, since aiida-quantumespresso plugin set this based on the number of atoms of the structure , so for big systems the step of the widget will affect this differently ... |
|
Good point! Anyway, we provide a small value so the user can control it. Otherwise, it's impossible for the user to control it using a mouse. |
I have an idea, just let me try something |
When doing a simulation with a system containing a higher number of atom , the convergence thresholds are quite lose.
For simulations using aiidalab-qe-vibroscopy this is impontant.
