Update README.md

README.md

@@ -2,12 +2,24 @@
 language:
 - en
 base_model:
-- unsloth/llama-3-8b-Instruct
 - unsloth/Meta-Llama-3.1-8B-Instruct
-- unsloth/Llama-3.2-3B-Instruct
 pipeline_tag: question-answering
 license: apache-2.0
 datasets:
 - NoCanGo/WaifuDB
 ---
-A project to create a chatbot that is knowledgeable in anime since the base models are severely lacking in that department.
+A project to create a chatbot that is knowledgeable in anime since the base models are severely lacking in that department.
+
+WaifuGPT3: Llama 3 8B
+WaifuGPT3.1: Llama 3.1 8B
+WaifuGPT3.2: Llama 3.2 3B
+
+Version 3 has less censorship and similar performance to 3.1.
+Version 3.1 is newer with similar quality than 3.
+Version 3.2 is less accurate but gives much quicker answers, perfect for an android device.
+
+Recommended temperature 0.5 for an accurate answer.
+Recommended temperature 0.7 for a creative answer.
+
+Usage example: Who would win a fight between Sakura Haruno and Tsunade? Expected answer example: Both Sakura Haruno and Tsunade are known for their physical strength but Tsunade's experience might give her a small edge.
+Usage example: Answer me as Tsunade to: What do you think about Sakura Haruno? Expected answer example: Heh Sakura is a skilled kunoichi but a little too fragile if you ask me.