Update README.md

README.md

@@ -1,16 +1,15 @@
 # Generate train data for follow model
 
-This repo will generate data of `Follow-Lang/set.mm` in huggingface.
+This repo will generate data of `Follow-Lang/set.mm.label` in huggingface.
 
 ## Format
 
 - The data is located in datasets/train.
-- Each line is formatted as: Cost(s) Cost(a) Cost(s')\t s a s' .
-- Cost(s) = max(Cost(a), 1+Cost(s')).
-- The maximum word length is 2048.
-- All vocabulary words are listed in datasets/words.txt.
+- Each line is formatted as: `s label arguments`.
+- The maximum word length is 1024.
+- All vocabulary words are listed in `words.txt` based on `Follow-Lang/set.mm.proof`.
 - The data was generated with a depth of 2.
 
 If you need additional data, feel free to reach out.
 
-This version improves readability and flow while maintaining the original meaning.
+This version improves readability and flow while maintaining the original meaning.