NeuralQuantumState.jl
NQS Settings
NeuralQuantumState.NETSETTINGS
— TypeNETSETTINGS
Generate settings by specifying values for non-default fields. Parsed with Parameters.jl
.
Arguments
modelname::String
: name of model (seesrc/weights.jl
).test_iter::Int
: Number of independent consecutive Carleo/Troyer runs.repetitions::Int
: Number of wavefunction optimizations.n::Int
: Number of visible neurons.α::Int
: Hidden unit density (No. hidden / No. visible).γ_init::Float
: Learning parameter (default:0.05
).γ_decay::Float
: Decay of learning parameter (default:1.0
).mc_trials::Int
: Number of variational derivatives PER repetition PER CPU.mc_steps::Int
: Number of MH steps per sample (default:n
).regulator::Float
: Regularization parameter of the S-matrix.init_therm_steps::Int
: Burn-in steps on first repetition.therm_steps::Int
: Burn-in on other repetitions.nflips::Int
: Number of flips to be performed per mc_trial (default:1
).pbc::Bool
: Periodic boundary conditions for lattice (default:true
).mag0::Bool
: Sample only in sector ∑s_z = 0 (default:false
).mfree::Bool
: Matrix-free inversion (default:true
). Overridden tofalse
. wheniterative_inverse = false
.iterative_inverse::Bool
: Invert S-matrix iteratively (default:true
). Assumed true whenmfree = true
.calc_stat::Bool
: Calculate single-site statistics and correlations after last repetition.stat_samples::Int
: No. of samples PER CPU for statistics calculation.save_figures::Bool
: Save statistics in figure (default:true
).writetofile::Bool
: Write optimized NQS parameters, energy[, statistics] to a file.use_meter::Bool
: Use a ProgressMeter bar to visualize progress (default:true
).
Examples
julia> NetSettings = NETSETTINGS(
modelname = "U_afh",
repetitions =1000,
n = 6,
α = 3,
mc_trials = 500,
init_therm_steps = 100,
therm_steps = 50,
stat_samples = 2000)
NETSETTINGS{String,Int64,Float64,Bool}
modelname: String "U_afh"
test_iter: Int64 1
repetitions: Int64 1000
n: Int64 6
α: Int64 3
m: Int64 18
dim_rbm: Int64 132
γ_init: Float64 0.05
γ_decay: Float64 1.0
mc_trials: Int64 500
mc_steps: Int64 6
regulator: Float64 0.1
init_therm_steps: Int64 100
therm_steps: Int64 50
nflips: Int64 1
pbc: Bool true
mag0: Bool false
mfree: Bool true
iterative_inverse: Bool true
calc_stat: Bool true
stat_samples: Int64 5000
save_figures: Bool true
writetofile: Bool true
use_meter: Bool true
Function
NeuralQuantumState.run_NQS
— Methodrun_NQS(NetSettings)
Arguments
- `NetSettings`: The settings struct containing all possible network settings.