Update README.md

README.md

@@ -28,6 +28,13 @@ BakLLaVA 2 is cooking with a significantly larger (commercially viable) dataset
 
 ![image/png](https://cdn-uploads.huggingface.co/production/uploads/64b7e345f92b20f7a38bf47a/qdYubrBmF7ztAHgdfkkwG.png)
 
+# Training dataset
+
+-558K filtered image-text pairs from LAION/CC/SBU, captioned by BLIP.
+-158K GPT-generated multimodal instruction-following data.
+-450K academic-task-oriented VQA data mixture.
+-40K ShareGPT data.
+-Additional private data (permissive)