Compatibility: jVMC v. 1.2.4
Main code for overlap calculation is in overlap.py. There is also a custom exact sampler in custom_exact_sampler.py which supports calculation in
chunks to avoid out-of-memory errors for large systems in exact overlap calculations.
Examples of usage are in examples/.
$ cd examples; python time_evol_1D_TFIM-ARNN.py --help
usage: time_evol_1D_TFIM-ARNN.py [-h] [--n_steps N_STEPS] [--n_Samples N_SAMPLES] [--hiddenSize HIDDENSIZE] [--depth DEPTH] [--timeStep TIMESTEP] [--batchSize BATCHSIZE] [--no_overlap_exact]
[--numChunks NUMCHUNKS] [--numSamples_MC_overlap NUMSAMPLES_MC_OVERLAP] [--verbose] [--numFidelityBins NUMFIDELITYBINS] [--fidelityBinsMin FIDELITYBINSMIN]
[--fidelityBinsMax FIDELITYBINSMAX]
L g seed
Computes time evolution of random NQS with 1D TFIM Hamiltonian. Calculates `numFidelityBins` fidelity and overlap estimates, each within a single bin, where bins range from `fidelityBinsMin` to `fidelityBinsMax`.
Uses autoregressive Ansatz with multi-layer RNN.
positional arguments:
L System size.
g Amplitude of sigma_x in the Hamiltonian.
seed Random seed.
optional arguments:
-h, --help show this help message and exit
--n_steps N_STEPS Number of time steps. (default: 2000)
--n_Samples N_SAMPLES
Number of MC samples in the time evolution by SR. (default: 200)
--hiddenSize HIDDENSIZE
Hidden size of RNN. (default: 2)
--depth DEPTH Vertical depth of RNN. (default: 2)
--timeStep TIMESTEP Time step of Euler integrator. (default: 0.0005)
--batchSize BATCHSIZE
Batch size for time integrator. Use power of 2. (default: 8192)
--no_overlap_exact Do not calculate the exact overlap and always return crude MC estimate. (default: False)
--numChunks NUMCHUNKS
Number of chunks of the full exact basis. Use even number. (default: 1)
--numSamples_MC_overlap NUMSAMPLES_MC_OVERLAP
Number of samples in the crude MC overlap estimation. (default: 512)
--verbose Print additional output. (default: False)
--numFidelityBins NUMFIDELITYBINS
Number of fidelity bins. (default: 10)
--fidelityBinsMin FIDELITYBINSMIN
Fidelity bins minimal value. (default: 0.0)
--fidelityBinsMax FIDELITYBINSMAX
Fidelity bins maximal value. (default: 1.0)
$ cd examples/; python time_evol_1D_TFIM-CpxRBM.py --help
usage: time_evol_1D_TFIM-CpxRBM.py [-h] [--n_steps N_STEPS] [--n_Samples N_SAMPLES] [--hiddenSize HIDDENSIZE] [--bias] [--timeStep TIMESTEP] [--batchSize BATCHSIZE] [--numChains NUMCHAINS]
[--sweepSteps SWEEPSTEPS] [--thermalizationSweeps THERMALIZATIONSWEEPS] [--no_overlap_exact] [--numChunks NUMCHUNKS] [--numSamples_MC_overlap NUMSAMPLES_MC_OVERLAP]
[--verbose] [--numFidelityBins NUMFIDELITYBINS] [--fidelityBinsMin FIDELITYBINSMIN] [--fidelityBinsMax FIDELITYBINSMAX]
L g seed
Computes time evolution of random NQS with 1D TFIM Hamiltonian. Calculates `numFidelityBins` fidelity and overlap estimates, each within a single bin, where bins range from `fidelityBinsMin` to `fidelityBinsMax`.
Uses RBM Ansatz with complex weights.
positional arguments:
L System size.
g Amplitude of sigma_x in the Hamiltonian.
seed Random seed.
optional arguments:
-h, --help show this help message and exit
--n_steps N_STEPS Number of time steps. (default: 2000)
--n_Samples N_SAMPLES
Number of MC samples in the time evolution by SR. (default: 200)
--hiddenSize HIDDENSIZE
Number of hidden units. (default: 2)
--bias Use bias. (default: False)
--timeStep TIMESTEP Time step of Euler integrator. (default: 0.0005)
--batchSize BATCHSIZE
Batch size for time integrator. (default: 8192)
--numChains NUMCHAINS
Number of MCMC chains run in parallel. (default: 100)
--sweepSteps SWEEPSTEPS
Number of MCMC steps for a sweep (in units of system size L) (default: 1)
--thermalizationSweeps THERMALIZATIONSWEEPS
Number of MCMC sweeps used for initial thermalization (burn-in). (default: 25)
--no_overlap_exact Do not calculate the exact overlap and always return crude MC estimate. (default: False)
--numChunks NUMCHUNKS
Number of chunks of the full exact basis. Use even number. (default: 1)
--numSamples_MC_overlap NUMSAMPLES_MC_OVERLAP
Number of samples in the crude MC overlap estimation. (default: 512)
--verbose Print additional output. (default: False)
--numFidelityBins NUMFIDELITYBINS
Number of fidelity bins. (default: 10)
--fidelityBinsMin FIDELITYBINSMIN
Fidelity bins minimal value. (default: 0.0)
--fidelityBinsMax FIDELITYBINSMAX
Fidelity bins maximal value. (default: 1.0)