Update app.py

app.py

@@ -42,14 +42,9 @@ def cached_vision_process(image, max_crops, num_tokens):
         return image_features.to(DEVICE, dtype=DTYPE)
 
 @spaces.GPU(duration=20)
-def answer_question(image, question, max_crops, num_tokens):
+def answer_question(image, question, max_crops, num_tokens, sample, temperature, top_k):
     prompt = f"""<|im_start|>user
 <image>
-{question}<|im_end|>
-<|im_start|>assistant
-"""
-    replace_prompt = f"""<|im_start|>user
-
 {question}<|im_end|>
 <|im_start|>assistant
 """
@@ -65,6 +60,9 @@ def answer_question(image, question, max_crops, num_tokens):
         "use_cache": True,
         "eos_token_id": processor.tokenizer.eos_token_id,
         "pad_token_id": processor.tokenizer.eos_token_id,
+        "temperature": temperature,
+        "sample": sample,
+        "top_k": top_k,
     }
     thread = Thread(target=model.generate, kwargs=generation_kwargs)
     thread.start()
@@ -101,8 +99,12 @@ with gr.Blocks() as demo:
         with gr.Row():
             img = gr.Image(type="pil", label="Upload or Drag an Image")
             output = gr.TextArea(label="Answer")
+        with gr.Row():
+            sample = gr.Checkbox(label="Sample", value=False)
+            temperature = gr.Slider(minimum=0, maximum=1, step=0.1, value=0, label="Temperature")
+            top_k = gr.Slider(minimum=0, maximum=50, step=1, value=0, label="Top-K")
 
-    submit.click(answer_question, [img, prompt, max_crops, num_tokens], output)
-    prompt.submit(answer_question, [img, prompt, max_crops, num_tokens], output)
+    submit.click(answer_question, [img, prompt, max_crops, num_tokens, sample, temperature, top_k], output)
+    prompt.submit(answer_question, [img, prompt, max_crops, num_tokens, sample, temperature, top_k], output)
 
-demo.queue().launch(debug=True, share=True)
+demo.queue().launch(debug=True)