Skip to content

Commit

Permalink
[doc] added doc strings and cleaned up parameter doc
Browse files Browse the repository at this point in the history
  • Loading branch information
merkelm committed Feb 27, 2024
1 parent e1c83a0 commit 3108ef8
Show file tree
Hide file tree
Showing 4 changed files with 57 additions and 20 deletions.
2 changes: 2 additions & 0 deletions python/solid_dmft/io_tools/dict_to_h5.py
Original file line number Diff line number Diff line change
Expand Up @@ -16,8 +16,10 @@ def _iteratively_replace_none(to_write, replace_from, replace_with):
return to_write

def prep_params_for_h5(dict_to_write):
""" Replace all NoneType with a string 'none' to be able to write to h5. """
return _iteratively_replace_none(deepcopy(dict_to_write), None, 'none')

# Not sure if the reverse route is actually needed
def prep_params_from_h5(dict_to_read):
""" Replace all 'none' strings with NoneType to parse the dict coming from h5. """
return _iteratively_replace_none(deepcopy(dict_to_read), 'none', None)
58 changes: 38 additions & 20 deletions python/solid_dmft/io_tools/documentation.txt
Original file line number Diff line number Diff line change
Expand Up @@ -67,6 +67,7 @@ h_field_it : int, default = 0
number of iterations the magnetic field is kept on
h_int_basis : string, default = 'triqs'
cubic basis convention to compute the interaction U matrix

* 'triqs'
* 'vasp' (equivalent to 'triqs')
* 'wien2k'
Expand All @@ -88,7 +89,8 @@ h_int_type : string, mandatory
h5_save_freq : int, default = 5
how often is the output saved to the h5 archive
J : float or list of float, mandatory
J values for impurities if only one value is given, the same J is assumed for all impurities
J interaction value. If it is a float, the same J is assumed for all impurities,
otherwise as a list a different J can be specified per impurity.
jobname : str, default = 'dmft_dir'
the output directory for one-shot calculations
load_sigma : bool, default = False
Expand All @@ -110,8 +112,8 @@ mu_gap_gb2_threshold : float, default = None
of MaxEnt's lattice spectral function to put the chemical potential
into the middle of the gap. Does not work if system completely full
or empty, mu mixing is not applied to it. Recommended value 0.01.
mu_gap_occ_deviation : float, default = none
Only used if mu_gap_gb2_threshold != none. Sets additional criterion
mu_gap_occ_deviation : float, default = None
Only used if mu_gap_gb2_threshold != None. Sets additional criterion
for finding the middle of the gap through occupation deviation to
avoid getting stuck in an insulating state with wrong occupation.
mu_initial_guess : float, default = None
Expand All @@ -137,11 +139,11 @@ n_iter_dmft_first : int, default = 10
n_iter_dmft_per : int, default = 2
number of iterations per dmft step in CSC calculations
n_iw : int, default = 1025
number of Matsubara frequencies
number of Matsubara frequencies for the imaginary-frequency grid
n_tau : int, default = 10001
number of imaginary time points
number of imaginary time points for the imaginary-time grid
n_w : int, default = 5001
number of real frequency points
number of real frequency points for the real-frequency grid
noise_level_initial_sigma : float, default = 0.0
spread of Gaussian noise applied to the initial Sigma
occ_conv_crit : float, default = -1.0
Expand All @@ -164,22 +166,29 @@ sampling_iterations : int, default = 0
seedname : str, mandatory
seedname for h5 archive with DMFT input and output
set_rot : string, default = None
use density_mat_dft to diagonalize occupations = 'den'
use hloc_dft to diagonalize occupations = 'hloc'
Local orbital rotations added by solid_dmft

* None: keep the rotations stored in the h5 archive
* 'den' use the DFT occupations density_mat_dft for diagonalization
* 'hloc': use the DFT local Hamiltonian hloc_dft for diagonalization
sigma_conv_crit : float, default = -1.0
stop the calculation if sum_w 1/(w^0.6) ||Sigma-Sigma_prev|| is smaller than threshold
sigma_mix : float, default = 1.0
careful: Sigma mixing can break orbital symmetries, use G0 mixing
careful: Sigma mixing can break orbital symmetries, use G0 mixing.
mixing sigma with previous iteration sigma for better convergency. 1.0 means no mixing
store_solver : bool, default = False
store the whole solver object under DMFT_input in h5 archive
whether to store the whole solver object under DMFT_input in h5 archive
U : float or list of float, mandatory
U values for impurities if only one value is given, the same U is assumed for all impurities
U interaction value. If it is a float, the same U is assumed for all impurities,
otherwise as a list a different U can be specified per impurity.
U_prime : float or list of floats, default = None
U prime values for impurities if only one value is given, the same U prime is assumed for all impurities
only used if h_int_type is kanamori
U prime interaction value.
Only used for impurities where general.h_int_type is kanamori.
If it is a float, the same U prime is assumed for all impurities,
otherwise as a list a different U prime can be specified per impurity.
For None, the default of U prime = U-2J is used.
w_range : list of int, default = [-10, 10]
Contains w_min and w_max
Minimal and maximal range of the real-frequency grid

[ solver ]
------------
Expand Down Expand Up @@ -224,9 +233,12 @@ max_time : int, default = -1
measure_chi_insertions : int, default = 100
number of insertation for measurement of chi
measure_chi : str, default = None
TODO: update
measure the dynamic spin suszeptibility chi(sz,sz(tau))
measure the dynamic suszeptibility of an operator O, chi(O,O(tau))
triqs.github.io/cthyb/unstable/guide/dynamic_susceptibility_notebook.html
Possible values for this flag are:

* 'SzSz': spin susceptibility
* 'NN': occupation susceptibility
measure_density_matrix : bool, default = False
measures the impurity density matrix and sets also
use_norm_as_weight to true
Expand Down Expand Up @@ -409,8 +421,14 @@ dft_exec : string, default = 'vasp_std'
command for the DFT executable
mpi_env : string, default = 'default'
selection for mpi env for DFT / VASP in default this will only call VASP as mpirun -np n_cores_dft dft_exec
Possible values:

* 'default'
* 'openmpi'
* 'openmpi-intra'
* 'mpich'
n_cores : int, None
number of cores for the DFT code (VASP)
number of cores for the DFT code (VASP). Mandatory for charge-self-consistent calculations
n_iter : int, default = 4
only needed for VASP. Number of DFT iterations to feed the DMFT
charge density into DFT, which generally takes multiple Davidson steps.
Expand Down Expand Up @@ -445,15 +463,15 @@ dc_nominal : bool, default = False
dc_orb_shift : list of float, default = None
extra potential shift per orbital per impurity added to the DC
dc_J : float or list of float, default = general.J
J values for DC determination if only one value is given, the same J is assumed for all impurities
J values for DC determination. If only one value is given, the same J is assumed for all impurities
dc_U : float or list of float, default = general.U
U values for DC determination if only one value is given, the same U is assumed for all impurities
U values for DC determination. If only one value is given, the same U is assumed for all impurities
map_solver_struct : list of dict, default = None
Additional manual mapping of the solver block structure, applied
after the block structure finder for each impurity.
Give exactly one dict per ineq impurity.
see also triqs.github.io/dft_tools/latest/_python_api/triqs_dft_tools.block_structure.BlockStructure.map_gf_struct_solver.html
mapped_solver_struct_degeneracies : list, default = none
mapped_solver_struct_degeneracies : list, default = None
Degeneracies applied when using map_solver_struct, for each impurity.
If not given and map_solver_struct is used, no symmetrization will happen.
pick_solver_struct : list of dict, default = None
Expand Down
15 changes: 15 additions & 0 deletions python/solid_dmft/io_tools/postproc_toml_dict.py
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,21 @@ def merge_config_with_default(cfg_inp, cfg_def, match_key={}):
- Mandatory inputs for all calculations indicated by "<no default>"
- None indicated by "<none>". Also works inside lists
- References within the dictionary indicated by "<section.key>"
Parameters
----------
cfg_inp : dict
The input config dict
cfg_def : dict
The default config dict
match_key : dict, optional
A dictionary that contains section/key pairs to map entries in listed sections
between the input and default config.
Returns
-------
dict
The merged config dict
"""

# Check restrictions and makes sure that config and default are of type FullConfig
Expand Down
2 changes: 2 additions & 0 deletions python/solid_dmft/io_tools/verify_input_params.py
Original file line number Diff line number Diff line change
Expand Up @@ -85,12 +85,14 @@ def _verify_input_params_advanced(params: FullConfig) -> None:
pass

def verify_before_dmft_cycle(params: FullConfig) -> None:
""" All checks of params that can be done before dmft_cycle is called. """
_verify_input_params_general(params)
_verify_input_params_dft(params)
_verify_input_params_solver(params)
_verify_input_params_advanced(params)

def verify_h5_dependent(sum_k, solver_type_per_imp, general_params):
""" All checks of params that depend on the h5 file content that is stored in the SumkDFT object. """
# Incompatabilities for SO coupling
if sum_k.SO == 1 and general_params['magnetic'] and general_params['afm_order']:
raise ValueError('AFM order not supported with SO coupling')
Expand Down

0 comments on commit 3108ef8

Please sign in to comment.