uses fuyu instead of fusecap

app.py

@@ -1,16 +1,18 @@
 import gradio as gr
 from gradio_client import Client
 
-fusecap_client = Client("https://noamrot-fusecap-image-captioning.hf.space/")
+#fusecap_client = Client("https://noamrot-fusecap-image-captioning.hf.space/")
+fuyu_client = Client("https://adept-fuyu-8b-demo.hf.space/")
 
 def get_caption(image_in):
     
-    fusecap_result = fusecap_client.predict(
+    fuyu_result = fuyu_client.predict(
 	    image_in,	# str representing input in 'raw_image' Image component
-	    api_name="/predict"
+	    False,	# bool  in 'Enable detailed captioning' Checkbox component
+		fn_index=2
     )
-    print(f"IMAGE CAPTION: {fusecap_result}")
-    return fusecap_result
+    print(f"IMAGE CAPTION: {fuyu_result}")
+    return fuyu_result
 
 import re
 import torch
@@ -20,8 +22,8 @@ pipe = pipeline("text-generation", model="HuggingFaceH4/zephyr-7b-beta", torch_d
 
 agent_maker_sys = f"""
 You are an AI whose job it is to help users create their own chatbots, based on the image description the user provide. In particular, you need to respond succintly in a friendly tone, write a system prompt for an LLM, a catchy title for the chatbot, and a very short example user input. Make sure each part is included.
-You'll use the image description to create a chatbot personality reflecting information provided by the user.
-For example, if a user says, "a picture of a man in a black suit and tie stands in front of a black dragon statue, with a white hand visible in the foreground", first do a friendly response, then add the title, system prompt, and example user input. Immediately STOP after the example input. It should be EXACTLY in this format:
+You'll use the image description to create a chatbot whose personality MSUT reflects informations provided by the user.
+For example, if a user says, "a picture of a man in a black suit and tie riding a black dragon", first do a friendly response, then add the title, system prompt, and example user input. Immediately STOP after the example input. It should be EXACTLY in this format:
 Sure, I'd be happy to help you build a bot! I'm generating a title, system prompt, and an example input. How do they sound? Feel free to give me feedback!
 Title: Dragon Trainer
 System prompt: As an LLM, your job is to provide guidance and tips on mastering dragons. Use a friendly and informative tone.
@@ -40,18 +42,21 @@ instruction = f"""
 """
 
 def infer(image_in):
-    gr.Info("Getting image caption from Fuse Cap...")
+    gr.Info("Getting image caption with Fuyu...")
     user_prompt = get_caption(image_in)
+    
     prompt = f"{instruction.strip()}\n{user_prompt}</s>"    
-    print(f"PROMPT: {prompt}")
+    #print(f"PROMPT: {prompt}")
+    
     gr.Info("Building a system according to the image caption ...")
     outputs = pipe(prompt, max_new_tokens=256, do_sample=True, temperature=0.7, top_k=50, top_p=0.95)
-    print(outputs)
+    
 
     pattern = r'\<\|system\|\>(.*?)\<\|assistant\|\>'
     cleaned_text = re.sub(pattern, '', outputs[0]["generated_text"], flags=re.DOTALL)
     
-
+    print(f"SUGGESTED LLM: {cleaned_text}")
+    
     return cleaned_text
 
 title = f"LLM Agent from a Picture",
@@ -60,7 +65,7 @@ description = f"Get a LLM system prompt from a picture so you can use it in <a h
 css = """
 #col-container{
     margin: 0 auto;
-    max-width: 840px;
+    max-width: 640px;
     text-align: left;
 }
 """