fix typo in readme

README.md

@@ -35,7 +35,7 @@ This was created for research purposes to investigate the use of embeddings obta
 
 See the [dataset card](https://huggingface.co/datasets/rchan26/english_char_split) for the exact train/validation/test splits of English words.
 
-In our training rum, we obtained a final loss of 0.0958.
+In our training run, we obtained a final validation loss of 0.0958.
 
 ## Training procedure