Changed clean.py to output to validation.csv

Changed the output of the validation dataset to validation.csv (formerly evaluation.csv) in order to align with HuggingFace expected splits.

clean.py

@@ -147,7 +147,7 @@ train_df = train_df[:train_size]
 # print("Saving to disk...")
 with open("train.csv", "w") as f:
     train_df.to_csv(f)
-with open("evaluation.csv", "w") as f:
+with open("validation.csv", "w") as f:
     eval_df.to_csv(f)
 with open("test.csv", "w") as f:
     test_df.to_csv(f)