Update README.md

README.md

@@ -1 +1 @@
-This contains 31 checkpoints from training a 8 layer GPT on chess games. If interested, one could examine how concepts such as board state understanding or skill level understanding emerge throughout the training process.
+This contains 31 checkpoints from training a 8 layer GPT on stockfish generated chess games. If interested, one could examine how concepts such as board state understanding emerge throughout the training process. NOTE: This will not work with probes for Lichess skill classification.