Update README.md

README.md

@@ -18,6 +18,26 @@ pipeline_tag: visual-document-retrieval
 
 ![bg](https://cdn-uploads.huggingface.co/production/uploads/6720a87e392e9cea0187fde6/nRa7iE30dqCUHGblnK8GQ.png)
 
+## 🚨 Migration guide to official `transformers` modeling
+
+ModernVBERT is part of `transformers` since v5.3.0!
+
+This becomes the official modeling for the model.
+
+If you have a model based on the initial modeling, you can adapt the weigths to the new format supported by `transformers`.
+We provide a conversion script in this repo, you simply need to run:
+
+```bash
+python convert_model_weights.py \
+  /path/to/legacy-modernvbert \
+  /path/to/converted-modernvbert \
+```
+
+The converted model:
+- merges the split token embedding tables
+- rewrites `model.connector.modality_projection.proj.weight`
+- nests `model.vision_model...` under `model.vision_model.vision_model...`
+
 ## Model
 This is the model card for `modernvbert`. 
 
@@ -100,22 +120,6 @@ predicted_token = tokenizer.decode(predicted_token_id)
 print("Predicted token:", predicted_token) # Predicted token: painting
 ```
 
-## Migration guide to official `transformers` modeling
-
-If you have a model based on the initial modeling, you will need to adapt the weigths to the new format supported by `transformers`.
-We provide a conversion script in this repo, you simply need to run:
-
-```bash
-python convert_model_weights.py \
-  /path/to/legacy-modernvbert \
-  /path/to/converted-modernvbert \
-```
-
-The converted model:
-- merges the split token embedding tables
-- rewrites `model.connector.modality_projection.proj.weight`
-- nests `model.vision_model...` under `model.vision_model.vision_model...`
-
 ## Evaluation