Update app.py

app.py

@@ -12,8 +12,50 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 import os
 password1=os.environ["password"]
 model_id = "peterpeter8585/ai2"
+gr.Interface(
+    respond,
+    inputs=[gr.MultimodalTextbox(file_types=["image"], show_label=False)],
+    outputs="text",
+    title="IDEFICS2-8B DPO",
+    description="Try IDEFICS2-8B fine-tuned using direct preference optimization (DPO) in this demo. Learn more about vision language model DPO integration of TRL [here](https://huggingface.co/blog/dpo_vlm).",
+    examples=[
+        {"text": "What is the type of flower in the image and what insect is on it?", "files": ["./bee.jpg"]},
+        {"text": "Describe the image", "files": ["./howl.jpg"]},
+    ],
+).launch()def respond1(
+    message,
+    history: list[tuple[str, str]],
+    system_message,
+    max_tokens,
+    temperature,
+    top_p,
+    password
+):
+    if password==password1:
+        messages = [{"role": "system", "content": "Your name is Chatchat.And your creator of you is Sung Yoon.In Korean, it is 정성윤.These are the instructions for you:"+system_message}]
 
-def respond1(multimodal_input, history: list[tuple[str, str]],system_message,max_tokens,temperature,top_p,password):
+        for val in history:
+            if val[0]:
+                messages.append({"role": "user", "content": val[0]})
+            if val[1]:
+                messages.append({"role": "assistant", "content": val[1]})
+
+        messages.append({"role": "user", "content": message})
+
+        response = ""
+
+        for message in client.chat_completion(
+            messages,
+            max_tokens=max_tokens,
+            stream=True,
+            temperature=temperature,
+            top_p=top_p,
+        ):
+            token = message.choices[0].delta.content
+
+            response += token
+            yield response
+def respond0(multimodal_input, history: list[tuple[str, str]],system_message,max_tokens,temperature,top_p,password):
     if password==password1:
         
         images = multimodal_input["files"]
@@ -430,9 +472,19 @@ ae= gr.ChatInterface(
         
     ],
 )
+a7=gr.Interface(
+    respond0,
+    inputs=[gr.MultimodalTextbox(file_types=["image"], show_label=False)],
+    outputs="text",
+    title="IDEFICS2-8B DPO",
+    description="Try IDEFICS2-8B fine-tuned using direct preference optimization (DPO) in this demo. Learn more about vision language model DPO integration of TRL [here](https://huggingface.co/blog/dpo_vlm).",
+    examples=[
+        {"text": "What is the type of flower in the image and what insect is on it?", "files": ["./bee.jpg"]},
+        {"text": "Describe the image", "files": ["./howl.jpg"]},
+    ],
+)
 aa=gr.ChatInterface(
     respond1,
-    textbox=gr.MultimodalTextbox(file_types=["image"], show_label=False),
     chatbot=chatbot3,
     additional_inputs=[
         gr.Textbox(value="You are a helpful assistant.", label="System message",  interactive=True),
@@ -482,5 +534,5 @@ ab= gr.ChatInterface(
 )
 if __name__ == "__main__":
     with gr.Blocks(theme="gstaff/xkcd") as ai:
-        gr.TabbedInterface([aa, ac, ab, ae, aaaa,demo2], ["gpt4(Password needed)", "gpt4(only for programming)", "gpt4(only for medical questions)", "gpt4(only for food recommendations)", "gpt4(only for law questions)","image create"])
+        gr.TabbedInterface([aa, ac, ab, ae, aaaa,demo2, a7], ["gpt4(Password needed)", "gpt4(only for programming)", "gpt4(only for medical questions)", "gpt4(only for food recommendations)", "gpt4(only for law questions)","image create", "multimodal"])
         ai.launch(share=True)