Merge pull request #724 from AIStream-Peelout/fix_bugs_series_id

Fix bugs series

flood_forecast/evaluator.py

@@ -367,6 +367,8 @@ def handle_later_ev(model, df_train_and_test, end_tensor, params, csv_test_loade
     if num_prediction_samples is not None:
         model.model.train()  # sets mode to train so the dropout layers will be touched
         assert num_prediction_samples > 0
+        if csv_test_loader.__class__.__name__ == "SeriesIDTestLoader":
+            raise NotImplementedError("SeriesIDTestLoader not yet supported for predictions.")
         prediction_samples = generate_prediction_samples(
             model,
             df_train_and_test,

flood_forecast/preprocessing/pytorch_loaders.py

@@ -230,11 +230,12 @@ def __getitem__(self, idx: int) -> Tuple[Dict, Dict]:
             targ_list = {}
             for va in self.listed_vals:
                 # We need to exclude the index column on one end and the series id column on the other
-                t = torch.Tensor(va.iloc[idx: self.forecast_history + idx].values)[:, 1:-1]
-                print(t.shape)
+
                 targ_start_idx = idx + self.forecast_history
                 idx2 = va[self.series_id_col].iloc[0]
-                targ = torch.Tensor(va.iloc[targ_start_idx: targ_start_idx + self.forecast_length].to_numpy())[:, 1:-1]
+                va_returned = va[va.columns.difference([self.series_id_col], sort=False)]
+                t = torch.Tensor(va_returned.iloc[idx: self.forecast_history + idx].values)[:, 1:]
+                targ = torch.Tensor(va_returned.iloc[targ_start_idx: targ_start_idx + self.forecast_length].to_numpy())[:, 1:] # noqa
                 src_list[self.unique_dict[idx2]] = t
                 targ_list[self.unique_dict[idx2]] = targ
             return src_list, targ_list
@@ -249,7 +250,7 @@ def __len__(self) -> int:
         if self.return_all_series:
             return len(self.listed_vals[0]) - self.forecast_history - self.forecast_length - 1
         else:
-            raise NotImplementedError("Current code only supports returning all the series at each iteration")
+            raise NotImplementedError("Current code only supports returning all the series at once at each iteration")
 
 
 class CSVTestLoader(CSVDataLoader):
@@ -667,11 +668,11 @@ class SeriesIDTestLoader(CSVSeriesIDLoader):
     def __init__(self, series_id_col: str, main_params: dict, return_method: str, forecast_total=336, return_all=True):
         """_summary_
 
-        :param series_id_col: _de
+        :param series_id_col: The column that contains the series_id
         :type series_id_col: str
-        :param main_params: _description_
+        :param main_params: The core params used to instantiate the CSVSeriesIDLoader
         :type main_params: dict
-        :param return_method: _description_
+        :param return_method: _description_D
         :type return_method: str
         :param return_all: _description_, defaults to True
         :type return_all: bool, optional

flood_forecast/temporal_decoding.py

@@ -7,7 +7,7 @@ def decoding_function(model, src: torch.Tensor, trg: torch.Tensor, forecast_leng
     """This function is responsible for decoding models that use `TemporalLoader` data. The basic logic of this
     function is as follows. The data to the encoder (e.g. src) is not modified at each step of the decoding process.
     Instead only the data to the decoder (e.g. the masked trg) is changed when forecasting max_len > forecast_length.
-    New data is appended (forecast_len == 2) (decoder_seq==10) (max==20) (20 (8)->2 First 8 should
+    New data is appended (forecast_len == 2) (decoder_seq==10) (max==20) (20 (8)->2 First 8 should be the same).
 
     :param model: The PyTorch time series forecasting model that you want to use forecasting on.
     :type model: `torch.nn.Module`
@@ -28,7 +28,7 @@ def decoding_function(model, src: torch.Tensor, trg: torch.Tensor, forecast_leng
     :type unknown_cols_st: List[str]
     :param decoder_seq_len: The length of the sequence passed into the decoder
     :type decoder_seq_len: int
-    :param max_len: The total number of time steps to forecast
+    :param max_len: The total number of time steps to forecas
     :type max_len: int
     :return: The forecasted values of shape (batch_size, max_len, n_targets)
     :rtype: torch.Tensor

flood_forecast/transformer_xl/cross_former.py

@@ -49,7 +49,7 @@ def __init__(
         :param baseline: _description_, defaults to False
         :type baseline: bool, optional
         :param device: _description_, defaults to torch.device("cuda:0")
-        :type device: _type_, optional
+        :type device: str, optional
         """
         super(Crossformer, self).__init__()
         self.data_dim = n_time_series

setup.py

@@ -16,7 +16,7 @@
 
 setup(
     name='flood_forecast',
-    version='1.000000dev',
+    version='1.00dev',
     packages=[
         'flood_forecast',
         'flood_forecast.transformer_xl',