l_init = mydevice.load(l_init)
q_label = mydevice.load(q_label)
p_label = mydevice.load(p_label)
return q_init,p_init,q_label,p_label,l_init
def print_dict(name,thisdict):
print(name,'dict ============== ')
for key,value in thisdict.items(): print(key,':',value)
if __name__=='__main__':
_ = mydevice()
_ = system_logs(mydevice)
system_logs.print_start_logs()
torch.set_default_dtype(torch.float64)
#print('checking pbc method ... wait a minute ')
#check_pbc()
torch.manual_seed(23841)
traindict = {"loadfile" : None, # to load previously trained model
"nn_mode" : 'hf', # 'hf' or 'ff' predict hamiltonian or predict force
"force_clip": 5, # clamp the force within a range
"grad_clip" : 10, # clamp the gradient for neural net parameters
"tau_long" : 0.1, # the large time step
"n_chain" : 1, # number of times to do integration before cal the loss
"ngrids" : 6, # for multibody interactions
"b" : 0.2, # grid lattice constant for multibody interactions
# python ML_trainer.py ../data/gen_by_ML/basename/MD_config.dict ../data/gen_by_ML/basename/ML_config.dict
argv = sys.argv
MDjson_file = argv[1]
MLjson_file = argv[2]
MD_parameters.load_dict(MDjson_file)
ML_parameters.load_dict(MLjson_file)
seed = ML_parameters.seed
torch.manual_seed(seed)
torch.set_default_dtype(torch.float64)
_ = system_logs(mydevice) # HK
system_logs.print_start_logs() # HK
# io varaiables
train_filename = ML_parameters.train_filename
val_filename = ML_parameters.valid_filename # read the same data
test_filename = val_filename
train_pts = ML_parameters.train_pts
val_pts = ML_parameters.valid_pts
test_pts = val_pts
qp_weight = ML_parameters.qp_weight
Lambda = ML_parameters.Lambda
clip_value = ML_parameters.clip_value
# crash checker variables
rthrsh0 = MD_parameters.rthrsh0
pthrsh0 = MD_parameters.pthrsh0
