Update README.md

README.md

@@ -20,7 +20,7 @@ It was trained on 32x32 terrain images from 265 melee maps including original Bl
 
 To run the demo with the ability to generate map files directly/with more coherence, use this notebook on Colab:
 
-https://colab.research.google.com/github/wdcqc/WaveFunctionDiffusion/blob/remaster/WaveFunctionDiffusion_Demo.ipynb
+https://colab.research.google.com/github/wdcqc/WaveFunctionDiffusion/blob/remaster/colab/WaveFunctionDiffusion_Demo.ipynb
 
 In addition to Dreambooth, a custom VAE model (`AutoencoderTile`) is trained to encode and decode the latents to/from tileset probabilities ("waves") and then generated as Starcraft maps.