load_weights will fail if shape mismatch (#10266)

Fix for #10265

keras/engine/saving.py

@@ -987,16 +987,24 @@ def load_weights_from_hdf5_group_by_name(f, layers, skip_mismatch=False,
                                      ' element(s).')
             # Set values.
             for i in range(len(weight_values)):
-                if skip_mismatch:
-                    if K.int_shape(symbolic_weights[i]) != weight_values[i].shape:
+                if K.int_shape(symbolic_weights[i]) != weight_values[i].shape:
+                    if skip_mismatch:
                         warnings.warn('Skipping loading of weights for layer {}'.format(layer.name) +
                                       ' due to mismatch in shape' +
                                       ' ({} vs {}).'.format(
                                           symbolic_weights[i].shape,
                                           weight_values[i].shape))
                         continue
-
-                weight_value_tuples.append((symbolic_weights[i],
-                                            weight_values[i]))
+                    else:
+                        raise ValueError('Layer #' + str(k) +
+                                         ' (named "' + layer.name +
+                                         '"), weight ' +
+                                         str(symbolic_weights[i]) +
+                                         ' has shape {}'.format(K.int_shape(symbolic_weights[i])) +
+                                         ', but the saved weight has shape ' +
+                                         str(weight_values[i].shape) + '.')
+                else:
+                    weight_value_tuples.append((symbolic_weights[i],
+                                                weight_values[i]))
 
     K.batch_set_value(weight_value_tuples)