Improve hparams saving and loading

pytorch_forecasting/models/base_model.py

@@ -244,7 +244,8 @@ def __init__(
         if not hasattr(self, "optimizer"):  # callables are removed from hyperparameters, so better to save them
             self.optimizer = self.hparams.optimizer
 
-        # delete everything from hparams that cannot be serialized with yaml dump
+        # delete everything from hparams that cannot be serialized with yaml.dump
+        # which is particularly important for tensorboard logging
         hparams_to_delete = []
         for k, v in self.hparams.items():
             try:
@@ -960,6 +961,8 @@ def on_save_checkpoint(self, checkpoint: Dict[str, Any]) -> None:
         checkpoint["hparams_name"] = "kwargs"
         # save specials
         checkpoint[self.CHECKPOINT_HYPER_PARAMS_SPECIAL_KEY] = {k: getattr(self, k) for k in self.hparams_special}
+        # add special hparams them back to save the hparams correctly for checkpoint
+        checkpoint[self.CHECKPOINT_HYPER_PARAMS_KEY].update(checkpoint[self.CHECKPOINT_HYPER_PARAMS_SPECIAL_KEY])
 
     @property
     def target_names(self) -> List[str]:

pytorch_forecasting/models/nbeats/sub_modules.py

@@ -26,7 +26,7 @@ def linspace(backcast_length: int, forecast_length: int, centered: bool = False)
         norm = backcast_length + forecast_length
         start = 0
         stop = backcast_length + forecast_length - 1
-    lin_space = np.linspace(start / norm, stop / norm, backcast_length + forecast_length, dtype=np.float6432)
+    lin_space = np.linspace(start / norm, stop / norm, backcast_length + forecast_length, dtype=np.float32)
     b_ls = lin_space[:backcast_length]
     f_ls = lin_space[backcast_length:]
     return b_ls, f_ls