diff --git a/neural_lam/constants.py b/neural_lam/constants.py
index 4b18081d..4b9ec009 100644
--- a/neural_lam/constants.py
+++ b/neural_lam/constants.py
@@ -179,7 +179,7 @@
 # Plotting
 FIG_SIZE = (15, 10)
 EXAMPLE_FILE = "data/cosmo/samples/train/data.zarr"
-EVAL_DATETIMES = ["2020050400"]  # prev_prev timestep (t-2)
+EVAL_DATETIMES = ["2015112800"]  # prev_prev timestep (t-2)
 EVAL_PLOT_VARS = ["T_2M"]
 STORE_EXAMPLE_DATA = True
 SELECTED_PROJ = ccrs.PlateCarree()
diff --git a/offline.py b/offline.py
index 52496bc9..6651d8bc 100644
--- a/offline.py
+++ b/offline.py
@@ -49,7 +49,7 @@ def offline_plotting():
     parser.add_argument(
         "--variable_to_plot",
         type=str,
-        default="TQV",
+        default="T_2M",
         help="Variable to plot in short format",
     )
 
@@ -86,7 +86,7 @@ def offline_plotting():
     start_time = pd.to_datetime(start_time_str, format="%Y%m%d%H")
 
     # Output the prediction time range
-    time_range = len(predictions[1, :, 1, 1])  # Number of time steps
+    time_range = len(predictions[0, :, 0, 0])  # Number of time steps
 
     # Calculate end time by adding the total duration of the time steps
     end_time = start_time + pd.Timedelta(hours=2 * time_range)