Update README.md

README.md

@@ -38,36 +38,3 @@ configs:
     path: data/train-*
 ---
 
-
-·▄▄▄▪   ▐ ▄ ▄▄▄ .▄▄▌ ▐ ▄▌▄▄▄ .▄▄▄▄·   
-▐▄▄·██ •█▌▐█▀▄.▀·██· █▌▐█▀▄.▀·▐█ ▀█▪   ╱|、🇰🇷  
-██▪ ▐█·▐█▐▐▌▐▀▀▪▄██▪▐█▐▐▌▐▀▀▪▄▐█▀▀█▄  (` - 7  
-██▌.▐█▌██▐█▌▐█▄▄▌▐█▌██▐█▌▐█▄▄▌██▄▪▐█  |、⁻〵  
-▀▀▀ ▀▀▀▀▀ █▪ ▀▀▀  ▀▀▀▀ ▀▪ ▀▀▀ ·▀▀▀▀   じしˍ,)ノ  
-.▄▄▄   ▄▄▄ .·▄▄▄▄  ▄• ▄▌  ▄ •▄▄▄▄     version 2.1.0  
-▐▀•▀█  ▀▄.▀·██▪ ██ █▪██▌  █▌▄▌▀▄ █·   FineWeb2-Edu-Korean  
-·▄▄▄▀  ▐▀▀▪▄▐█· ▐█▌█▌▐█▌  ▐▀▀▄▐▀▀▄   
-▄▄▄▪   ▐█▄▄▌██. ██ ▐█▄█▌  ▐█.█▐█•█▌  
-▀▀▀▀▀   ▀▀▀ ▀▀▀▀▀•  ▀▀▀   ·▀  .▀  ▀  
-  
-============================================================================  
-📚 The FineWeb2-Edu-Korean dataset is an educational web page filtered version  
-of the kor_Hang subset from the Fineweb2 dataset released by HuggingFaceFW.  
-  
-USAGE  
-Depending on your purpose, select only data above a certain score and use it  
-for Korean language model pretraining.  
-  
-LIMITATIONS / TODOS  
-- The filtering process requires significant computing power.  
-- The classifier model was trained and evaluated by a third party;  
-    it was not trained or evaluated directly by this project.  
-- The training quality has not been compared with the dataset before filtering.  
-  
-  
-ACKNOWLEDGEMENTS  
-minpeter :b  
-HuggingfaceFW, the original dataset (Fineweb2) authors,  
-  and for providing the idea for edu data filtering  
-devngho, provided the classifier model (ko_edu_classifier_v2_nlpai-lab_KoE5)  
-allura-org, inspired this README.txt from the dumbo project