Update README.md

README.md

@@ -26,7 +26,7 @@ To maintain efficiency and accessibility:
 - Vision and language components are frozen
 - **Only the adapter layer is trained**
 - **Supervised training approach, treating adapter training as model finetuning**(Houlsby et al. (2019)'s work on MLP adapters for transfer learning)
-- Was be trained on accessible hardware (T4 or P100 GPUs)
+- Was trained on accessible hardware (T4 or P100 GPUs)
 
 ## Demo 
 I couldn't afford the GPU prices here so it runs quite slooowww of CPU 🤧