def test_runoff_example(self, tmp_path):
cfg = Config(Path("tests/testconfigs/config_runoff.yml"))
create_and_assign_temp_run_path_to_config(cfg, tmp_path)
cfg._cfg["data_path"] = Path("data/ALL_dynamic_ds.nc")
cfg._cfg["static_data_path"] = Path("data/camels_static.nc")
cfg._cfg["static_inputs"] = ["p_mean", "pet_mean", "area", "gauge_elev"]
cfg._cfg["n_epochs"] = 3
ds, static = load_data(cfg)
# select subset of 3 basins
basins = [1001, 2001, 2002]
ds = ds.sel(station_id=basins)
static = static.sel(station_id=basins)
trainer = Trainer(cfg, ds, static_data=static)
self.check_loaded_data(
cfg,
trainer,
data=ds.sel(time=slice(cfg.train_start_date, cfg.train_end_date)),
)
losses = trainer.train_and_validate()
tester = Tester(cfg, ds, static_data=static)
preds = tester.run_test()
return losses, preds
def main(args):
reader = load_data(args)
# Run oracle
scores = []
validator = KFoldValidator(num_folds=args.folds, verbose=args.verbose)
for item in reader.scored_items():
best_params, best_params_score = validator.train([item])
scores.append(best_params_score)
if args.verbose:
print(item.name, best_params, best_params_score, sep='\t')
if args.summarize:
print('all', '---', statistics.mean(scores), sep='\t')
def test_kenya_vci_example(self, tmp_path):
cfg = Config(Path("tests/testconfigs/config.yml"))
create_and_assign_temp_run_path_to_config(cfg, tmp_path)
cfg._cfg["data_path"] = Path("data/kenya.nc")
cfg._cfg["n_epochs"] = 3
ds, static = load_data(cfg)
trainer = Trainer(cfg, ds, static_data=static)
self.check_loaded_data(
cfg,
trainer,
data=ds.sel(time=slice(cfg.train_start_date, cfg.train_end_date)),
)
losses = trainer.train_and_validate()
tester = Tester(cfg, ds, static_data=static)
preds = tester.run_test()
return losses, preds
def test_file_list(self):
data = load_data(["fibers.trk", "data.nii"])
self.assertEqual(data, ["fibers.trk", "data.nii"])
def test_single_file(self):
data = load_data(["fibers.trk"])
self.assertEqual(data, ["fibers.trk"])
def begin(self):
super(FiberTrackingHook, self).begin()
self.test_set = run.load_data(self.test_set)
dirs[int(raw_input('enter your choice: '))])
print("Loading params from: " + path)
args = np.load(os.path.join(path, 'args.pkl'))
return args, path
def init_model():
model = VRAE(args.rnn_size,
args.rnn_size,
args.n_features,
args.latent_size,
num_drivers,
batch_size=args.batch_size)
model.create_gradientfunctions(x_train, t_train, x_valid, t_valid)
return model
def load_epoch(model, e):
model.load_parameters(os.path.join(path, str(e)))
def forward(model, batch_num=0):
return model.encoder(x_train[batch_num * model.batch_size:(batch_num + 1) *
model.batch_size].transpose(1, 0, 2))
args, path = load_params()
x_train, t_train, x_valid, t_valid = load_data(args)
num_drivers = np.max(t_train) + 1
model = init_model()