Update app.py

app.py

@@ -11,6 +11,7 @@ model, vis_processors, _ = load_model_and_preprocess(
     model_type=model_type,
     is_eval=True,
     device=device,
+    dtype=torch.float16 
 )
 
 def infer(image, prompt, min_len, max_len, beam_size, len_penalty, repetition_penalty, top_p, decoding_method):
@@ -53,73 +54,74 @@ with gr.Blocks(theme=theme, analytics_enabled=False,css=css) as demo:
             The demo is based on the official <a href="https://github.com/salesforce/LAVIS/tree/main/projects/instructblip" style="text-decoration: underline;" target="_blank"> Github </a> implementation
             """
         )
-    with gr.Column(scale=3):
-        image_input = gr.Image(type="pil")
-        prompt_textbox = gr.Textbox(label="Prompt:", placeholder="prompt", lines=2)
-        output = gr.Textbox(label="Output")
-        submit = gr.Button("Run", variant="primary")
+    with gr.Row():
+        with gr.Column(scale=3):
+            image_input = gr.Image(type="pil")
+            prompt_textbox = gr.Textbox(label="Prompt:", placeholder="prompt", lines=2)
+            output = gr.Textbox(label="Output")
+            submit = gr.Button("Run", variant="primary")
 
-    with gr.Column(scale=1):
-        min_len = gr.Slider(
-            minimum=1,
-            maximum=50,
-            value=1,
-            step=1,
-            interactive=True,
-            label="Min Length",
-        )
-    
-        max_len = gr.Slider(
-            minimum=10,
-            maximum=500,
-            value=250,
-            step=5,
-            interactive=True,
-            label="Max Length",
-        )
-    
-        sampling = gr.Radio(
-            choices=["Beam search", "Nucleus sampling"],
-            value="Beam search",
-            label="Text Decoding Method",
-            interactive=True,
-        )
-    
-        top_p = gr.Slider(
-            minimum=0.5,
-            maximum=1.0,
-            value=0.9,
-            step=0.1,
-            interactive=True,
-            label="Top p",
-        )
-    
-        beam_size = gr.Slider(
-            minimum=1,
-            maximum=10,
-            value=5,
-            step=1,
-            interactive=True,
-            label="Beam Size",
-        )
-    
-        len_penalty = gr.Slider(
-            minimum=-1,
-            maximum=2,
-            value=1,
-            step=0.2,
-            interactive=True,
-            label="Length Penalty",
-        )
-    
-        repetition_penalty = gr.Slider(
-            minimum=-1,
-            maximum=3,
-            value=1,
-            step=0.2,
-            interactive=True,
-            label="Repetition Penalty",
-        )
+        with gr.Column(scale=1):
+            min_len = gr.Slider(
+                minimum=1,
+                maximum=50,
+                value=1,
+                step=1,
+                interactive=True,
+                label="Min Length",
+            )
+        
+            max_len = gr.Slider(
+                minimum=10,
+                maximum=500,
+                value=250,
+                step=5,
+                interactive=True,
+                label="Max Length",
+            )
+        
+            sampling = gr.Radio(
+                choices=["Beam search", "Nucleus sampling"],
+                value="Beam search",
+                label="Text Decoding Method",
+                interactive=True,
+            )
+        
+            top_p = gr.Slider(
+                minimum=0.5,
+                maximum=1.0,
+                value=0.9,
+                step=0.1,
+                interactive=True,
+                label="Top p",
+            )
+        
+            beam_size = gr.Slider(
+                minimum=1,
+                maximum=10,
+                value=5,
+                step=1,
+                interactive=True,
+                label="Beam Size",
+            )
+        
+            len_penalty = gr.Slider(
+                minimum=-1,
+                maximum=2,
+                value=1,
+                step=0.2,
+                interactive=True,
+                label="Length Penalty",
+            )
+        
+            repetition_penalty = gr.Slider(
+                minimum=-1,
+                maximum=3,
+                value=1,
+                step=0.2,
+                interactive=True,
+                label="Repetition Penalty",
+            )
     
     submit.click(infer, inputs=[image_input, prompt_textbox, min_len, max_len, beam_size, len_penalty, repetition_penalty, top_p, sampling], outputs=[output])