Update README.md

README.md

@@ -69,18 +69,18 @@ This technique avoids training unnecessarily low-performng weights that can turn
 
 Axolotl was used for training and dataset tokenization. 
 
-#### Preprocessing [optional]
+#### Preprocessing 
 
 Dataset was formatted in ShareGpt format for the purposes of using with Axolotl, in conversational format.
 
 #### Training Hyperparameters
 
-lora_r: 64
-lora_alpha: 16
-lora_dropout: 0.05
-gradient_accumulation_steps: 4
-micro_batch_size: 1
-num_epochs: 3
-optimizer: adamw_bnb_8bit
-lr_scheduler: cosine
-learning_rate: 0.00025
+- lora_r: 64
+- lora_alpha: 16
+- lora_dropout: 0.05
+- gradient_accumulation_steps: 4
+- micro_batch_size: 1
+- num_epochs: 3
+- optimizer: adamw_bnb_8bit
+- lr_scheduler: cosine
+- learning_rate: 0.00025