Settings of a YamboCalculation

The settings Dict that we provide as input of a YamboCalculation is a fundamental quantity that can decide the logic of pre-yambo executable action that are accomplished by the plugin. To understand the possible actions, we need to explain how the plugin works in standard cases.

The plugin currently supports, four type of logic to run a calculation:

  1. p2y from a NSCF

    this will just run a p2y calculation to create the Yambo SAVE database, before to run, the nscf save folder is copied to the new remote directory. It is triggered by using as parent calculation an NSCF run with the quantumespresso.pw plugin and by setting:

    inputs['settings'] = ParameterData(dict={'INITIALISE': True})
  2. yambo from a p2y

    triggered simply by using as parent calculation a p2y run with the yambo plugin. This will, by default, create a link of the SAVE directory contained in the p2y remote folder.

  3. p2y + yambo from a NSCF

    triggered by using as parent calculation an NSCF calculation run with the quantumespresso.pw plugin

  4. yambo from a (previous) yambo

    useful in particular for restarts, it is triggered by using as parent calculation a Yambo calculation run with the yambo plugin.

If you want also to link the output databases produced from the previous yambo calculation, you can set:

inputs['settings'] = ParameterData(dict={'RESTART_YAMBO': True})

in this way you can continue your calculation from the last point. These are the main logics of a typical YamboCalculation. There can be problems in the linking of the SAVE and output directories, so you can tell the plugin to make an hard copy of the folder of interest:

inputs['settings'] = ParameterData(dict={'COPY_SAVE': True})
inputs['settings'] = ParameterData(dict={'COPY_DBS': True})

so, a complete settings Dict will be:

inputs['settings'] = ParameterData(dict={'INITIALISE': True,
                                         'RESTART_YAMBO': True,
                                         'COPY_SAVE': True,
                                         'COPY_DBS': True,
                                         })

Primer on Yambo parallelizations

The computational effort done during a Yambo calculation requires an extensive and wise use of parallelization schemes on various quantities that are computed during the simulation. There are two ways to find an automatic parallelization scheme in the AiiDA-yambo plugin: to use predefined Yambo-core parallelization utilities or to use the parallelizer provided among the plugin. In any case, when yambo sees a parallelization problem before to start the real calculation, tries to use it default scheme. If this fails, the failed calculation will be corrected from the plugin built-in parallelizer.

default yambo parallelization:

just put, in the parameters dictionary, the instruction

'PAR_def_mode': "balanced"       # [PARALLEL] Default distribution mode ("balanced"/"memory"/"workload")

yambo-aiida parallelizer:

you can choose the roles to be parallelized between bands or kpoints, or both; this may modify your resources by fitting them to the dimensions of the simulation, but only changing mpi-openmpi balance or reducing the total number of processors if they are too much (example: you may want parallelize 100 bands with 150 CPUs -> reduce CPUs to 100).

To use the plugin parallelizer:

from aiida_yambo.utils.parallelism_finder import *

find_parallelism_qp(nodes, mpi_per_node, threads, bands, occupied=2, qp_corrected=2, kpoints = 1, \
                    what = ['bands'], last_qp = 2)

the input what is a list of what you want to parallelize:bands, kpoints, g. You have also to provide some useful information like the computational resources, the total number of bands that you have in the simulation, the occupied states, and so on. The output will be two dictionaries, the first for the parallelization part of the input parameters, the second one for the resources (nodes, mpi and threads).

Parsing results using built-in functions

You can parse data like quasiparticle corrections (single-levels or gaps) by using a built-in function that is used in the YamboConvergence workchain. This function is the calc_manager_aiida_yambo, and it is useful if you want to parse in a fast way results from a calculation. Let’s say that you want to extract the gap computed in a single YamboCalculation, and that you know the k-points and the bands corresponding to the gap, then you can use the calc_manager_aiida_yambo in this way:

from aiida_yambo.workflows.utils.helpers_aiida_yambo import *

x = calc_manager_aiida_yambo()
x.wfl_pk = 7720 #workflow(YamboWorkflow or YamboRestart) pk or calculation
x.take_quantities(backtrace = 1,what='gap',where = [(1,4,1,5)])

what=single-levels can be another options to parse qp corrections: in that case where=[(kv,bv),(kc,v)], where v is for valence and c is for conduction.