add init params

paddlex/cv/models/yolo_v3.py

@@ -144,7 +144,7 @@ class YOLOv3(BaseAPI):
             iou_aware_factor=self.iou_aware_factor,
             use_drop_block=self.use_drop_block,
             batch_size=self.train_batch_size,
-            max_shape=self.max_shape)
+            max_shape=self.max_shape if hasattr(self, 'max_shape') else 608)
         inputs = model.generate_inputs()
         model_out = model.build_net(inputs)
         outputs = OrderedDict([('bbox', model_out)])
@@ -253,6 +253,7 @@ class YOLOv3(BaseAPI):
                 if isinstance(transform, paddlex.det.transforms.Normalize):
                     transform.is_scale = False
         if self.use_iou_loss or self.use_iou_aware_loss:
+            self.init_params['train_batch_size'] = train_batch_size
             self.max_shape = 0
             for transform in train_dataset.transforms.transforms:
                 if isinstance(transform, paddlex.det.transforms.Resize):