[ENH] de-novo implementation of LTSFTransformer based on cure-lab research code base

docs/source/api_reference/forecasting.rst

@@ -430,6 +430,7 @@ Deep learning based forecasters
     LTSFLinearForecaster
     LTSFDLinearForecaster
     LTSFNLinearForecaster
+    LTSFTransformerForecaster
 
 .. currentmodule:: sktime.forecasting.hf_transformers_forecaster
 

sktime/forecasting/base/adapters/_pytorch.py

@@ -135,6 +135,7 @@ def _predict(self, X=None, fh=None):
         else:
             dataloader = self.build_pytorch_pred_dataloader(X, fh)
 
+        self.network.eval()
         y_pred = []
         for x, _ in dataloader:
             y_pred.append(self.network(x).detach())