Update README.md

README.md

@@ -5,3 +5,17 @@ license: afl-3.0
 ---
 LyricaLlama is a finetuned version of [meta-llama/Llama-2-7b-hf](https://huggingface.co/meta-llama/Llama-2-7b-hf) which was trained on publicly available English lyrics data.  
 The purpose of the model was to experiment with fintuning and creativity in LLM's.  Results were close to what you would expect from a Llama 7b model but subjectively outperforms the base model.
+
+Training Format:
+```
+### Instruction: You are a creative, world-famous expert lyricist. Write lyrics for a song, given just a title, artist name, possible genres, and any additional information provided.
+
+### Input:
+Write lyrics for a song titled "{song_name}" to be performed by {artist}[ using the following geners: {comma delimited list of genres}].
+
+### Response:
+{lyrics}
+```
+
+Note that genre's were optional during training, but recommended during generation
+