phase_space = phase_space.phase_space()
if crash_checker == 'yes':
if crash_thrsh == 'yes':
crsh_chker = check4particle_soft_crash(rthrsh0, pthrsh0, rthrsh,
pthrsh, crash_path)
else:
crsh_chker = check4particle_hard_crash(rthrsh, pthrsh, crash_path)
else:
crsh_chker = check4particle_crash_dummy(rthrsh0, pthrsh0, rthrsh,
pthrsh, crash_path)
linear_integrator_obj = linear_integrator(MD_parameters.integrator_method,
crsh_chker)
hamiltonian_obj = make_hamiltonian(hamiltonian_type, tau_long,
ML_parameters)
if hamiltonian_type != "noML": # use prediction for ML
chk_pt = checkpoint(hamiltonian_obj.get_netlist(
)) # opt = None, sch = None ; for test, don't need opt, sch
if load_model_file is not None: chk_pt.load_checkpoint(load_model_file)
hamiltonian_obj.eval()
hamiltonian_obj.requires_grad_false()
init_qpl, _, _ = data_io.read_trajectory_qpl(MC_init_config_filename)
# init_qp_bs.shape = [nsamples, 3=(q, p, boxsize), 1, nparticle, DIM]
init_q = torch.squeeze(init_qpl[:, 0, 0, :, :], dim=1)
# init_q.shape = [nsamples, nparticle, DIM]
init_p = torch.squeeze(init_qpl[:, 1, 0, :, :], dim=1)
crsh_chker = check4particle_hard_crash(rthrsh, pthrsh, crash_path)
else:
crsh_chker = check4particle_crash_dummy(rthrsh0, pthrsh0, rthrsh,
pthrsh, crash_path)
linear_integrator_obj = linear_integrator(MD_parameters.integrator_method,
crsh_chker)
# don't check crash to make input into fhnn
linear_integrator_dummy_obj = linear_integrator(
MD_parameters.integrator_method,
check4particle_crash_dummy(rthrsh0, pthrsh0, rthrsh, pthrsh,
crash_path))
hamiltonian_obj = make_hamiltonian(hamiltonian_type,
linear_integrator_dummy_obj, tau_short,
tau_long, ML_parameters)
if hamiltonian_type != "noML": # use crash
chk_pt = checkpoint(hamiltonian_obj.get_netlist(
)) # opt = None, sch = None ; for test, don't need opt, sch
if load_model_file is not None: chk_pt.load_checkpoint(load_model_file)
hamiltonian_obj.eval()
hamiltonian_obj.requires_grad_false()
init_qp, _, _, boxsize = data_io.read_trajectory_qp(
MC_init_config_filename)
# init_qp.shape = [nsamples, (q, p), 1, nparticle, DIM]
init_q = torch.squeeze(init_qp[:, 0, 0, :, :], dim=1)
# init_q.shape = [nsamples, nparticle, DIM]