Update README.md

README.md

@@ -7,15 +7,8 @@ datasets:
 - jtatman/medquad-medicalqa-wizdolalpaca-instruct
 - jtatman/medical_biological_instruction_format
 ---
-This now has interesting results with medically relevant information.
+This is an ongoing experiment in training and retraining boundaries.
 
+The model is currently overtrained and is purposely so to investigate the paths out of overtraining.
 
-Try the ChatML prompt style:
-
-<|im_start|>user\n
-What are the three causes of pneumonia?<|im_end|>\n
-<|im_start|>assistant\n
-
-The inference API has some weird effects, and probably amplifies things. 
-
-Better to try out through pipelines. 
+This is purely an experiment on depths and depravity of repetitive training. Don't bother messing around with it much.