Create README.md

this a random subset of 300k images from diffusion db, along with captions generated by blip2.
This data is used as generated images during training of the fakeinversion model.
This is not the same subset of images that was used by the original authors.

README.md

@@ -0,0 +1,5 @@
+---
+language:
+- en
+pretty_name: FakeInversion DifussionDB 300k - subset
+---