fix

README.md

@@ -87,7 +87,7 @@ The full fine-tuning process can be expensive, specially the bigger the model. A
 For SAM, I chose to use the LoRA adapters.
 
 ## Low-Rank Adaptation (LoRA)
-LoRA is an adapter that is using 2 matrices B and A. The 2 matrices have specific dimensions (input_size, r) and (r, input_size) . By specifying a rank r < input_size, we reduce the parameters size and try to capture the task with a small enough rank. By doing the dot product B*A, we get a matrix of shape (input_size, input_size) so no information is lost but the model will have learned a new representation through training.
+LoRA is an adapter that is using 2 matrices B and A. The 2 matrices have specific dimensions (input_size, r) and (r, input_size) . By specifying a rank r < input_size, we reduce the parameters size and try to capture the task with a small enough rank. The matrix product B*A gives a matrix of shape (input_size, input_size) so no information is lost but the model will have learned a new representation through training.
 
 For our application, we only need to initialize the matrices, freeze SAM and train the adapter so that the frozen model + LoRA learns to segment rings.