resolve merge issues

03b_model/model_config.yaml

@@ -11,6 +11,7 @@ inputs: ['discharge_01463500',
                 'wind_direction_delsjmet',
                 'wind_speed_delsjmet']#,'wind_speed_direction_delsjmet']                    
 
+
 #the sinks to be analyzed, this will likely be the salt front location
 target: 'saltfront7_weekly'
 replicates: 5
@@ -45,3 +46,4 @@ dropout: 0.1
 n_epochs: 500
 learn_rate: 0.001
 
+

03b_model/src/LSTMDA_torch.py

@@ -99,35 +99,34 @@ def mse_masked(y_true, y_pred):
     mse_loss = sum_squared_errors / num_y_true
     return mse_loss
 
-def extreme_loss(y_true, y_pred):
-    y_true_or_err = torch.where(
-        torch.isnan(y_true), torch.Tensor([-9999]) , y_true
-    )
-    #-0.0138 is approx river mile 70 scaled by mean and sd of the training set
-    low_zero_or_error = torch.where(
-        (y_true_or_err >= -0.0138) | (y_true_or_err < -999), torch.zeros_like(y_true_or_err), y_pred - y_true_or_err
-        )
-    num_low_y_true = torch.count_nonzero(
-        low_zero_or_error
-        )
-    high_zero_or_error = torch.where(
-        (y_true_or_err < -0.0138) , torch.zeros_like(y_true_or_err), y_pred - y_true_or_err
-        )
-    num_high_y_true = torch.count_nonzero(
-        high_zero_or_error
-        )
+# def extreme_loss(y_true, y_pred):
+#     y_true_or_err = torch.where(
+#         torch.isnan(y_true), torch.Tensor([-9999]) , y_true
+#     )
+#     #-0.0138 is approx river mile 70 scaled by mean and sd of the training set
+#     low_zero_or_error = torch.where(
+#         (y_true_or_err >= -0.0138) | (y_true_or_err < -999), torch.zeros_like(y_true_or_err), y_pred - y_true_or_err
+#         )
+#     num_low_y_true = torch.count_nonzero(
+#         low_zero_or_error
+#         )
+#     high_zero_or_error = torch.where(
+#         (y_true_or_err < -0.0138) , torch.zeros_like(y_true_or_err), y_pred - y_true_or_err
+#         )
+#     num_high_y_true = torch.count_nonzero(
+#         high_zero_or_error
+#         )
 
-    sum_squared_errors_low = torch.sum(torch.square(low_zero_or_error))
-    loss_low = torch.sqrt(sum_squared_errors_low / num_low_y_true)
-
-    #cube the errors where river mile is high
-    sum_cubed_errors_high = torch.sum(torch.pow(high_zero_or_error,4))
-    loss_high = (sum_cubed_errors_high / num_high_y_true)
+#     sum_squared_errors_low = torch.sum(torch.square(low_zero_or_error))
+#     loss_low = torch.sqrt(sum_squared_errors_low / num_low_y_true)
 
-    loss_hi_low = loss_low.add(loss_high)
+#     #cube the errors where river mile is high
+#     sum_cubed_errors_high = torch.sum(torch.pow(high_zero_or_error,4))
+#     loss_high = (sum_cubed_errors_high / num_high_y_true)
 
-    return loss_hi_low
+#     loss_hi_low = loss_low.add(loss_high)
 
+#     return loss_hi_low
 
 
 # def rmse_weighted(y_true, y_pred): # weighted by covariance matrix from DA; weights are concatonated onto y_true and need to separate out within function 

03b_model/src/run_model.py

@@ -149,8 +149,13 @@ def select_inputs_targets(inputs, target, train_start_date, test_end_date, out_d
     # else:
     #     mask = target_df_c['saltfront_daily'] < 54
     #     target_df_c.loc[mask,'saltfront_daily'] = np.nan
+<<<<<<< HEAD
     mask = target_df_c['saltfront7_weekly'] < 54
     target_df_c.loc[mask,'saltfront7_weekly'] = np.nan
+=======
+    mask = target_df_c['saltfront_daily'] < 54
+    target_df_c.loc[mask,'saltfront_daily'] = np.nan
+>>>>>>> ba69d4c077cc19a40e3c4275ae19fec7a9fc1163
 
     inputs_xarray = inputs_df.to_xarray()
     target_xarray = target_df_c.to_xarray()
@@ -359,7 +364,10 @@ def write_model_params(out_dir, run_id, inputs, n_epochs,
     f = open(os.path.join(dir,"model_param_output.txt"),"w+")
     f.write("Date: %s\r\n" % date.today().strftime("%b-%d-%Y"))
     f.write("Feature List: %s\r\n" % inputs_log)
+<<<<<<< HEAD
     f.write("Target: %s\r\n" % target)
+=======
+>>>>>>> ba69d4c077cc19a40e3c4275ae19fec7a9fc1163
     f.write("Include antecedant variable: %s\r\n" % inc_ante)
     f.write("Epochs: %d\r\n" % n_epochs)
     f.write("Learning rate: %f\r\n" % learn_rate)
@@ -617,6 +625,7 @@ def run_replicates(n_reps, prepped_model_io_data_file):
                              test_start_date, test_end_date)
 
         plot_save_predictions(predictions, out_dir, run_id)
+
 
 
 def test_hyperparameters():

Snakefile_fetch_munge

@@ -89,6 +89,3 @@ rule munge_noaa_nerrs:
         munge = importlib.import_module('02_munge.src.munge_noaa_nerrs')
         munge.munge_single_site_data('delsjmet')
 
-
-
-

environment.yaml

@@ -39,3 +39,7 @@ dependencies:
       - pypi_install
       - sciencebasepy
   - scikit-learn
+
+
+
+