Merge pull request #55 from EthanJamesLew/bugbix/sp

correct bad generate predictions with sampling period correction (#52)

autokoopman/autokoopman.py

@@ -377,7 +377,12 @@ def _sanitize_training_data(
     # convert the data to autokoopman trajectories
     if isinstance(training_data, TrajectoriesData):
         if not isinstance(training_data, UniformTimeTrajectoriesData):
+            print(f"resampling trajectories as they need to be uniform time (sampling period {sampling_period})")
             training_data = training_data.interp_uniform_time(sampling_period)
+        else:
+            if not np.isclose(training_data.sampling_period, sampling_period):
+                print(f"resampling trajectories because the sampling periods differ (original {training_data.sampling_period}, new {sampling_period})")
+                training_data = training_data.interp_uniform_time(sampling_period)
     else:
         # figure out how to add inputs
         training_iter = (

autokoopman/core/tuner.py

@@ -178,13 +178,18 @@ def generate_predictions(
         preds = {}
         # get the predictions
         for k, v in holdout_data._trajs.items():
+            # ugh, this is a hack--sampling period is meaningful if the system is discrete
+            # TODO: make the behavior of sampling period less problematic
+            _kwargs = {}
+            if hasattr(v, "sampling_period"):
+                _kwargs["sampling_period"] = v.sampling_period
             sivp_interp = trained_model.model.solve_ivp(
                 v.states[0],
                 (np.min(v.times), np.max(v.times)),
                 inputs=v.inputs,
                 teval=v.times,
+                **_kwargs
             )
-            # sivp_interp = sivp_interp.interp1d(v.times)
             preds[k] = sivp_interp
         return TrajectoriesData(preds)