iit

.github/workflows/pylinter.yml

@@ -18,6 +18,7 @@ jobs:
         with:
           source: "src"
           mypy-options: "--ignore-missing-imports --show-error-codes"
+          black-options: "--exclude **/*"
           #pylint-options: "--rcfile=setup.cfg"
           isort-options: "-w 100"
           django: true

src/experiments/covid_exp.py

@@ -6,6 +6,7 @@
 import matplotlib.pyplot as plt
 from src.models import rnn, lstm, bjrnn, cfrnn, copulaCPTS, dplstm, vanila_copula
 
+
 class COVIDDataset(torch.utils.data.Dataset):
     def __init__(self, X, Y, sequence_lengths):
         super(COVIDDataset, self).__init__()
@@ -20,64 +21,65 @@ def __getitem__(self, idx):
         return self.X[idx], self.Y[idx], self.sequence_lengths[idx]
 
 
+def experiment(train, valid, test, target_len=50, name="exp"):
+    rnn_model = rnn.rnn(
+        input_size=1, embedding_size=128, output_size=1, horizon=target_len
+    )
+    encdec_model = lstm.lstm_seq2seq(
+        input_size=1, output_size=1, embedding_size=128, target_len=target_len
+    )
+    models = [rnn_model]  # , encdec_model]
 
-def experiment(train, valid, test, target_len = 50,name='exp'):
-    rnn_model = rnn.rnn(input_size=1, embedding_size=128, output_size=1, horizon=target_len)
-    encdec_model = lstm.lstm_seq2seq(input_size=1, output_size=1 , embedding_size=128, target_len =target_len)
-    models = [rnn_model] #, encdec_model]
-
     x_train = train.X
     y_train = train.Y
-    
+
     for m in models:
-        m.train_model(x_train,y_train, n_epochs=200, batch_size=50)
-    
-    with open('./trained/models_%s.pkl'%name, 'wb') as f:
-        pickle.dump(models , f)
-        
+        m.train_model(x_train, y_train, n_epochs=200, batch_size=50)
+
+    with open("./trained/models_%s.pkl" % name, "wb") as f:
+        pickle.dump(models, f)
+
     x_cali = valid.X
     y_cali = valid.Y
-    
+
     UQ = {}
-
-
-    dprnn = dplstm.DPRNN(epochs=150,  input_size=1, output_size=1, n_steps=target_len, dropout_prob=0.2)
+
+    dprnn = dplstm.DPRNN(
+        epochs=150, input_size=1, output_size=1, n_steps=target_len, dropout_prob=0.2
+    )
     dprnn.fit(x_train, y_train)
-    UQ['dprnn'] = dprnn
-    
-    bj_class = bjrnn.bj_rnn(models[0],recursion_depth=15)
+    UQ["dprnn"] = dprnn
+
+    bj_class = bjrnn.bj_rnn(models[0], recursion_depth=15)
     bj_class.LOBO_residuals = np.expand_dims(bj_class.LOBO_residuals, axis=-1)
-    UQ['bjrnn'] = bj_class   
+    UQ["bjrnn"] = bj_class
 
     cf = cfrnn.CFRNN(models[0], x_cali, y_cali)
     cf.calibrate()
-    UQ['cfrnn'] = cf
-    
+    UQ["cfrnn"] = cf
+
     cf = cfrnn.CFRNN(models[1], x_cali, y_cali)
     cf.calibrate()
-    UQ['cf-EncDec'] = cf
-
+    UQ["cf-EncDec"] = cf
 
     copula = copulaCPTS.copulaCPTS(models[0], x_cali, y_cali)
     copula.calibrate()
-    UQ['copula-rnn'] = copula
+    UQ["copula-rnn"] = copula
 
     copula = copulaCPTS.copulaCPTS(models[1], x_cali, y_cali)
     copula.calibrate()
-    UQ['copula-EncDec'] = copula
-
+    UQ["copula-EncDec"] = copula
 
     vanilla = vanila_copula.vanila_copula(models[0], x_cali, y_cali)
     vanilla.calibrate()
-    UQ['vanila-copula'] = vanilla
+    UQ["vanila-copula"] = vanilla
 
     x_test = test.X
-    y_test = test.Y
+    y_test = test.Y 
 
     areas = {}
     coverages = {}
 
-
     epsilon_ls = np.linspace(0.05, 0.50, 10)
 
     for k, uqmethodß in UQ.items():
@@ -91,27 +93,32 @@ def experiment(train, valid, test, target_len = 50,name='exp'):
             coverage.append(uqmethod.calc_coverage_1d(box, pred, y_test))
         areas[k] = area
         coverages[k] = coverage
-    
-    with open('./trained/uq_%s.pkl'%name, 'wb') as f:
-        pickle.dump(UQ , f)
-    with open('./trained/results_%s.pkl'%name, 'wb') as f:
-        pickle.dump((areas, coverages) , f)
-    
+
+    with open("./trained/uq_%s.pkl" % name, "wb") as f:
+        pickle.dump(UQ, f)
+    with open("./trained/results_%s.pkl" % name, "wb") as f:
+        pickle.dump((areas, coverages), f)
+
     return areas, coverages, (models, UQ)
 
+
 def main():
 
     with open("processed_data/covid_conformal_scaled.pkl", "rb") as f:
         train_dataset, calibration_dataset, test_dataset = pickle.load(f)
 
     for i in range(3):
-        res = experiment(train_dataset, calibration_dataset, test_dataset, target_len=50, 
-                        name='covid_daily_vanilla'+str(i)) # target len 6
-        print('finished run ' + str(i))
+        res = experiment(
+            train_dataset,
+            calibration_dataset,
+            test_dataset,
+            target_len=50,
+            name="covid_daily_vanilla" + str(i),
+        )  # target len 6
+        print("finished run " + str(i))
 
         del res
 
 
-
-if __name__ == '__main__':
+if __name__ == "__main__":
     main()