get_ipython().run_line_magic('pinfo2', 'learn.validate')
learn.lr_find(start_lr=1e-6,end_lr=1e-2,no_grad_val = False,num_it=63)
learn.recorder.plot()
learn.fit_one_cycle(cyc_len=8, max_lr=2e-5,moms=(0.95, 0.85), div_factor=10.0, pct_start=0.3, wd=1e-7,no_grad_val=False)
flatten_model(learn.model)
learn.split(groupness)
learn.layer_groups
learn.freeze_to(1)
learn.lr_find(start_lr=1e-6,end_lr=1e-2,no_grad_val = False,num_it=63)
learn.recorder.plot()
learn.fit_one_cycle(cyc_len=16, max_lr=2e-4,moms=(0.95, 0.85), div_factor=400.0, pct_start=0.1, wd=1e-7,no_grad_val=False)
model.summary()
learn.save('surfsmall')
learn.sched.plot()
learn.loss_func = partial(MSEloss,kf=1,ke=0.01)
learn.lr_find(start_lr=1e-10,end_lr=1e-4,no_grad_val = False,num_it=300)
learn.recorder.plot()
learn.fit_one_cycle(cyc_len=6, max_lr=1e-6,moms=(0.95, 0.85), div_factor=10.0, pct_start=0.1, wd=1e-7,no_grad_val=False)
# else:
# learn.model.load_state_dict(torch.load(os.path.join(args.modelpath,sys.argv[1]+".pth"))['model'])
# print("weights were loaded")
#except:
# print("no weights were loaded")
#pdb.set_trace()
#with torch.autograd.detect_anomaly():
#learn.fit_one_cycle(cyc_len=int(sys.argv[3]), max_lr=lrs,moms=(0.95, 0.85), div_factor=50.0, pct_start=0.30, wd=wds,no_grad_val=False)
learn.fit_one_cycle(cyc_len=int(sys.argv[3]),
max_lr=lrs,
moms=(0.95, 0.85),
div_factor=150.0,
pct_start=0.05,
wd=wds,
no_grad_val=False)
if args.uncertainty_forces:
print(
learn.save("bulkVtrain3200" + "_uncertain_forces_" + str(sys.argv[3]),
return_path=True))
elif args.uncertainty:
print(
learn.save("bulkVtrain3200" + "_uncertain_" + str(sys.argv[3]) + "_" +
str(args.p) + "_" + str(args.var_coeff),
return_path=True))
else:
print(
learn.save("bulkVtrain3200" + "_" + str(sys.argv[3]),
return_path=True))
copyfile(args.split_path, split_path)
#from sklearn.model_selection import train_test_split
#train,test = train_test_split(df, test_size=0.20, random_state=42,stratify=df['Ebin'].values)
print(args.batch_size)
train_loader = schnetpack2.custom.data.AtomsLoader(data_train, batch_size=args.batch_size, sampler=RandomSampler(data_train),
num_workers=0*torch.cuda.device_count(), pin_memory=True)#9*torch.cuda.device_count()
val_loader = schnetpack2.custom.data.AtomsLoader(data_val, batch_size=args.batch_size, num_workers=9*torch.cuda.device_count(), pin_memory=True)#9*torch.cuda.device_count()
mean, stddev = train_loader.get_statistics('energy', False)
#mean, stddev = torch.tensor([-1.5115]), torch.tensor([1.2643])
#stddev = 29.6958
print(mean,stddev)
model = get_model(train_args, atomref=None, mean=torch.FloatTensor([mean]), stddev=torch.FloatTensor([stddev]),
train_loader=train_loader,
parallelize=args.parallel)
data = DataBunch( train_loader, val_loader,collate_fn=schnetpack2.custom.data.collate_atoms)
learn = Learner(data, model, model_dir=args.modelpath)
learn.purge()
learn.opt_func = Adam
learn.loss_func = partial(MSEloss,kf=1.0,ke=0.1)
learn.metrics=[Emetric,Fmetric]
#learn.load('trainbulkVsurf128epochs')
torch.cuda.empty_cache()
print(sys.argv)
learn.fit_one_cycle(cyc_len=int(sys.argv[3]), max_lr=float(sys.argv[4]),moms=(0.95, 0.85), div_factor=150.0, pct_start=0.05, wd=1e-2,no_grad_val=False)
print(learn.save( sys.argv[1], return_path=True))
learn.metrics = [
Emetric, Fmetric
] #DipoleMetric, ChargeMetric, CouplingMetric, CouplingMetricKaggle]
#for epoch in range(1, int(sys.argv[2]) + 1):
# with gpytorch.settings.use_toeplitz(True):
# train(epoch)
# test()
#learn.load('trainbulkVsurf128epochs')
torch.cuda.empty_cache()
#print(sys.argv)
#with gpytorch.settings.use_toeplitz(True):
learn.fit_one_cycle(cyc_len=int(sys.argv[2]),
max_lr=float(sys.argv[3]),
moms=(0.95, 0.85),
div_factor=500.0,
pct_start=0.05,
wd=1e-2,
no_grad_val=False)
#print(learn.save(os.path.join(args.modelpath,"PhysNet_{0}_{1}_{2}".format(sys.argv[1], sys.argv[2], sys.argv[3])), return_path=True))
#torch.save(model, os.path.join(args.modelpath,"DKL_weights_{0}_{1}_{2}".format(sys.argv[1], sys.argv[2], sys.argv[3])))
#if args.uncertainty_forces:
# print(learn.save( "bulkVtrain3200"+"_uncertain_forces_"+str(sys.argv[3]), return_path=True))
if args.num_inducing_points:
print(
learn.save("bulkVtrain3200" + "_induced_" + str(sys.argv[3]),
return_path=True))
else:
print(learn.save("bulkVtrain3200_" + str(sys.argv[3]), return_path=True))