Fix disagg preprocessing, score disagg baseline models

DLinearTrain.py

@@ -0,0 +1,36 @@
+
+
+
+# Specify D-Linear model
+model_dlinear = DLinear(
+  input_chunk_length = 30,
+  output_chunk_length = 15,
+  kernel_size = 25,
+  batch_size = 32,
+  n_epochs = 500,
+  model_name = "DLinearDisagg2.0",
+  log_tensorboard = True,
+  save_checkpoints = True,
+  show_warnings = True,
+  optimizer_kwargs = {"lr": 0.002},
+  lr_scheduler_cls = torch.optim.lr_scheduler.ReduceLROnPlateau,
+  lr_scheduler_kwargs = {"patience": 5},
+  pl_trainer_kwargs = {
+    "callbacks": [early_stopper, progress_bar, model_summary],
+    "accelerator": "gpu",
+    "devices": [0]
+    }
+)
+
+
+# Fit DLinear model
+model_dlinear.fit(
+  series = [y[:-45] for y in y_train_disagg],
+  future_covariates = [x[dlinear2_futcovars] for x in x_disagg],
+  past_covariates = [x[dlinear2_pastcovars] for x in x_disagg],
+  val_series = [y[-45:] for y in y_train_disagg],
+  val_future_covariates = [x[dlinear2_futcovars] for x in x_disagg],
+  val_past_covariates = [x[dlinear2_pastcovars] for x in x_disagg],
+  verbose = True,
+  num_loader_workers = 20
+)