def create(data_nm, dim=1):
data_nm = data_nm.lower()
if data_nm == "mnist":
if dim == 1:
return MNISTDataset.get_tf_dataset_1d()
elif data_nm == "era_dataset":
return EraDataset.get()
def load(data_nm, dim=1):
data_nm = data_nm.lower()
if data_nm == "mnist":
try:
return MNISTDataset.load(dim, True)
except Exception as e:
print("dimension is max 3")
return None, None
if data_nm == "cifar10":
if dim == 1:
return Cifar10Dataset.get_tf_dataset_1d()
if __name__ == '__main__':
parameters = {
## model parameters
"model_nm": "DNN-test",
"algorithm_type": "classifier",
"job_type": "learn",
## learning parameters
"global_step": "10",
"early_type": "none",
"min_step": "10",
"early_key": "accuracy",
"early_value": "0.98",
## algorithm parameters
"input_units": "784",
"output_units": "10",
"hidden_units": "100, 200, 100",
"dropout_prob": "0.1",
"optimizer_fn": "Adam",
"learning_rate": "0.01",
"initial_weight": "0.1",
"act_fn": "tanh",
}
dnn = DNN(parameters)
dnn.build()
from hps.dataset.MNISTDataset import MNISTDataset
ds_learn, ds_test = MNISTDataset.get_tf_dataset_1d()
dnn.learn(ds_learn)
print(dnn.predict(ds_test))