Add VQA

app_caption.py

@@ -15,7 +15,7 @@ def create_demo():
     with gr.Row():
         with gr.Column():
             image = gr.Image(label='Input', type='filepath')
-            model_name = gr.Dropdown(label='Model', choices=['Prismer-Base, Prismer-Large'], value='Prismer-Base')
+            model_name = gr.Dropdown(label='Model', choices=['Prismer-Base', 'Prismer-Large'], value='Prismer-Base')
             run_button = gr.Button('Run')
         with gr.Column(scale=1.5):
             caption = gr.Text(label='Model Prediction')

app_vqa.py

@@ -44,9 +44,9 @@ def create_demo():
     gr.Examples(examples=examples,
                 inputs=inputs,
                 outputs=outputs,
-                fn=model.run_vqa_model)
+                fn=model.run_vqa)
 
-    run_button.click(fn=model.run_vqa_model, inputs=inputs, outputs=outputs)
+    run_button.click(fn=model.run_vqa, inputs=inputs, outputs=outputs)
 
 
 if __name__ == '__main__':

prismer_model.py

@@ -145,7 +145,7 @@ class Model:
         test_loader = create_loader(test_dataset, batch_size=1, num_workers=4, train=False)
         experts, _ = next(iter(test_loader))
         question = pre_question(question)
-        answer = self.model(experts, question, train=False, inference='generate')
+        answer = self.model(experts, [question], train=False, inference='generate')
         answer = self.tokenizer(answer, max_length=30, padding='max_length', return_tensors='pt').input_ids
         answer = answer.to(experts['rgb'].device)[0]
         answer = self.tokenizer.decode(answer, skip_special_tokens=True)