docs: add forward declaration steps to example to avoid pickling error (docarray#1615)

Joan Fontanals · alexcg1 · web-flow · commit 29c2d23a6187 · 2023-06-02T13:55:20.000+02:00
Signed-off-by: Joan Fontanals Martinez &lt;joan.martinez@jina.ai&gt;
Signed-off-by: Joan Fontanals &lt;jfontanalsmartinez@gmail.com&gt;
Co-authored-by: Alex Cureton-Griffiths &lt;alexcg1@users.noreply.github.com&gt;
diff --git a/docs/how_to/multimodal_training_and_serving.md b/docs/how_to/multimodal_training_and_serving.md
@@ -135,6 +135,18 @@ class PairTextImage(BaseDoc):
     image: ImageDoc
 ```
 
+You then need to forward declare the following types. This will allow the objects to be properly pickled and unpickled.
+
+This will be unnecessary once [this issue](https://github.com/docarray/docarray/issues/1330) is resolved.
+
+```python
+from docarray import DocVec
+DocVec[Tokens]
+DocVec[TextDoc]
+DocVec[ImageDoc]
+DocVec[PairTextImage]
+```
+
 ### Create the dataset 
 
 In this section we will create a multimodal pytorch dataset around the Flick8k dataset using DocArray.
