# Configuration files that customize the default behaviour of non-linear searches.

# **PyAutoFit** supports the following MCMC algorithms:

# - Emcee: https://github.com/dfm/emcee / https://emcee.readthedocs.io/en/stable/
# - Zeus: https://github.com/minaskar/zeus / https://zeus-mcmc.readthedocs.io/en/latest/

# Settings in the [search] and [run] entries are specific to each nested algorithm and should be determined by
# consulting that MCMC method's own readthedocs.

Emcee:
  run:
    nsteps: 2000
  search:
    nwalkers: 50
  auto_correlations:
    change_threshold: 0.01          # The threshold value by which if the change in auto_correlations is below sampling will be terminated early.
    check_for_convergence: true     # Whether the auto-correlation lengths of the Emcee samples are checked to determine the stopping criteria. If `True`, Emcee may stop before nsteps are performed.
    check_size: 100                 # The length of the samples used to check the auto-correlation lengths (from the latest sample backwards).
    required_length: 50             # The length an auto_correlation chain must be for it to be used to evaluate whether its change threshold is sufficiently small to terminate sampling early.
  initialize:                       # The method used to generate where walkers are initialized in parameter space {prior | ball}.
    method: ball                    # priors: samples are initialized by randomly drawing from each parameter's prior. ball: samples are initialized by randomly drawing unit values from a narrow uniform distribution.
    ball_lower_limit: 0.49          # The lower limit of the uniform distribution unit values are drawn from when initializing walkers using the ball method.
    ball_upper_limit: 0.51          # The upper limit of the uniform distribution unit values are drawn from when initializing walkers using the ball method.
  parallel:
    number_of_cores: 1              # The number of cores the search is parallelized over by default, using Python multiprocessing.
  printing:
    silence: false                  # If True, the default print output of the non-linear search is silcened and not printed by the Python interpreter.
  prior_passer:
    sigma: 3.0                      # For non-linear search chaining and model prior passing, the sigma value of the inferred model parameter used as the sigma of the passed Gaussian prior.
    use_errors: true                # If True, the errors of the previous model's results are used when passing priors.
    use_widths: true                # If True the width of the model parameters defined in the priors config file are used.
  updates:
    iterations_per_update: 500      # The number of iterations of the non-linear search performed between every 'update', where an update performs tasks like outputting model.results.
    remove_state_files_at_end: true # Whether to remove the savestate of the seach (e.g. the Emcee hdf5 file) at the end to save hard-disk space (results are still stored as PyAutoFit pickles and loadable).
Zeus:
  run:
    check_walkers: true
    light_mode: false
    maxiter: 10000
    maxsteps: 10000
    mu: 1.0
    nsteps: 2000
    patience: 5
    shuffle_ensemble: true
    tolerance: 0.05
    tune: true
    vectorize: false
  search:
    nwalkers: 50
  auto_correlations:
    change_threshold: 0.01          # The threshold value by which if the change in auto_correlations is below sampling will be terminated early.
    check_for_convergence: true     # Whether the auto-correlation lengths of the Emcee samples are checked to determine the stopping criteria. If `True`, Emcee may stop before nsteps are performed.
    check_size: 100                 # The length of the samples used to check the auto-correlation lengths (from the latest sample backwards).
    required_length: 50             # The length an auto_correlation chain must be for it to be used to evaluate whether its change threshold is sufficiently small to terminate sampling early.
  initialize:                       # The method used to generate where walkers are initialized in parameter space {prior | ball}.
    method: ball                    # priors: samples are initialized by randomly drawing from each parameter's prior. ball: samples are initialized by randomly drawing unit values from a narrow uniform distribution.
    ball_lower_limit: 0.49          # The lower limit of the uniform distribution unit values are drawn from when initializing walkers using the ball method.
    ball_upper_limit: 0.51          # The upper limit of the uniform distribution unit values are drawn from when initializing walkers using the ball method.
  parallel:
    number_of_cores: 1              # The number of cores the search is parallelized over by default, using Python multiprocessing.
  printing:
    silence: false                  # If True, the default print output of the non-linear search is silenced and not printed by the Python interpreter.
  prior_passer:
    sigma: 3.0                      # For non-linear search chaining and model prior passing, the sigma value of the inferred model parameter used as the sigma of the passed Gaussian prior.
    use_errors: true                # If True, the errors of the previous model's results are used when passing priors.
    use_widths: true                # If True the width of the model parameters defined in the priors config file are used.
  updates:
    iterations_per_update: 500      # The number of iterations of the non-linear search performed between every 'update', where an update performs tasks like outputting model.results.
    remove_state_files_at_end: true # Whether to remove the savestate of the seach (e.g. the Emcee hdf5 file) at the end to save hard-disk space (results are still stored as PyAutoFit pickles and loadable).