Update app.py

app.py

@@ -6,29 +6,39 @@ import torch
 from PIL import Image
 from transformers import AutoProcessor, LlavaForConditionalGeneration
 from transformers import TextIteratorStreamer
-
+from datasets import load_dataset
 import spaces
+import pandas as pd
+
+rekaeval = "RekaAI/VibeEval"
+dataset = load_dataset(rekaeval, split="test")
+df = pd.DataFrame(dataset)
+df_markdown = df.copy()
+
+# Function to convert URL to HTML img tag
+def mediaurl_to_img_tag(url):
+    return f'<img src="{url}">'
+
+# Apply the function to the DataFrame column
+df_markdown['media_url'] = df_markdown['media_url'].apply(mediaurl_to_img_tag)
 
 
 PLACEHOLDER = """
 <div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
-   <img src="https://cdn-uploads.huggingface.co/production/uploads/64ccdc322e592905f922a06e/DDIW0kbWmdOQWwy4XMhwX.png" style="width: 80%; max-width: 550px; height: auto; opacity: 0.55;  "> 
-   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">LLaVA-Llama-3-8B</h1>
-   <p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">Llava-Llama-3-8b is a LLaVA model fine-tuned from Meta-Llama-3-8B-Instruct and CLIP-ViT-Large-patch14-336 with ShareGPT4V-PT and InternVL-SFT by XTuner</p>
+   <img src="https://avatars.githubusercontent.com/u/51063788?s=400&u=479ecc9d93d8a373b5c2e69ebe846f394811e94a&v=4)" style="width:40%" opacity="0.45">
+   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">LLaVA-Llama3-8B With REKA Vibe-Eval</h1>
+   <p style="font-size: 18px; margin-bottom: 2px; opacity: 0.65;">Test your Vision LLMs with new Vibe-Evals from REKA</p>
 </div>
 """
 
 
 model_id = "xtuner/llava-llama-3-8b-v1_1-transformers"
-
 processor = AutoProcessor.from_pretrained(model_id)
-
 model = LlavaForConditionalGeneration.from_pretrained(
     model_id,
     torch_dtype=torch.float16,
     low_cpu_mem_usage=True,
 )
-
 model.to("cuda:0")
 model.generation_config.eos_token_id = 128009
 
@@ -88,10 +98,10 @@ def bot_streaming(message, history):
 
 chatbot=gr.Chatbot(placeholder=PLACEHOLDER,scale=1)
 chat_input = gr.MultimodalTextbox(interactive=True, file_types=["image"], placeholder="Enter message or upload file...", show_label=False)
-with gr.Blocks(fill_height=True, ) as demo:
+tmp = '''with gr.Blocks(fill_height=True, ) as demo:
     gr.ChatInterface(
     fn=bot_streaming,
-    title="LLaVA Llama-3-8B",
+    title="Testing LLaVA-Llama3-8b with Reka's Vibe-Eval",
     examples=[{"text": "What is on the flower?", "files": ["./bee.jpg"]},
               {"text": "How to make this pastry?", "files": ["./baklava.png"]}],
     description="Try [LLaVA Llama-3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers). Upload an image and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error.",
@@ -99,7 +109,66 @@ with gr.Blocks(fill_height=True, ) as demo:
     multimodal=True,
     textbox=chat_input,
     chatbot=chatbot,
-    )
-
-demo.queue(api_open=False)
-demo.launch(show_api=False, share=False)
+    )'''
+
+with gr.Blocks() as demo:
+  with gr.Row():
+    with gr.Column():
+      gr.ChatInterface(
+            fn=bot_streaming,
+            title="Testing LLaVA-Llama3-8b with Reka's Vibe-Eval",
+            description="Try [LLaVA Llama-3-8B](https://huggingface.co/xtuner/llava-llama-3-8b-v1_1-transformers). Upload an image and start chatting about it, or simply try one of the examples below. If you don't upload an image, you will receive an error.",
+            stop_btn="Stop Generation",
+            multimodal=True,
+            textbox=chat_input,
+            chatbot=chatbot,
+        )
+    with gr.Column():
+      with gr.Row():
+        b1 = gr.Button("Previous")
+        b2 = gr.Button("Next")
+      reka = gr.Dataframe(value=df_markdown[0:5], datatype=['markdown', 'str'], wrap=False, interactive=False, height=500)
+      num_start = gr.Number(visible=False, value=0)
+      num_end = gr.Number(visible=False, value=4)
+    
+  chat_msg = chat_input.submit(add_message, [chatbot, chat_input], [chatbot, chat_input])
+  bot_msg = chat_msg.then(bot, chatbot, chatbot, api_name="bot_response")
+  bot_msg.then(lambda: gr.MultimodalTextbox(interactive=True), None, [chat_input])
+  chatbot.like(print_like_dislike, None, None)
+
+  def get_example(reka, evt: gr.SelectData):
+      print(f'evt.value = {evt.value}')
+      print(f'evt.index = {evt.index}')
+      x = evt.index[0]
+      image = df.iloc[x, 0]  
+      prompt = df.iloc[x, 1]
+      print(f'image = {image}')
+      print(f'prompt = {prompt}')
+      example = {"text": prompt, "files": [image]}
+      return example
+
+  def display_next(dataframe, end):
+    print(f'initial value of end = {end}')
+    start = (end  or dataframe.index[-1]) + 1
+    end = start + 4
+    df_images = df_markdown.loc[start:end]
+    print(f'returned value of end = {end}')
+    print(f'returned value of start = {start}')
+    return df_images, end, start, gr.Button(interactive=True)
+
+  def display_previous(dataframe, start):
+    print(f'initial value of start = {start}')
+    end = (start  or dataframe.index[-1]) 
+    start = end - 5
+    df_images = df_markdown.loc[start:end]
+    print(f'returned value of start = {start}')
+    print(f'returned value of end = {end}')
+    return df_images, end, start, gr.Button(interactive=False) if start==0 else gr.Button(interactive=True)
+
+  reka.select(get_example, reka, chat_input, show_progress="hidden")
+  b2.click(fn=display_next, inputs= [reka, num_end ], outputs=[reka, num_end, num_start, b1], api_name="next_rows", show_progress=False)
+  b1.click(fn=display_previous, inputs= [reka, num_start ], outputs=[reka, num_end, num_start, b1], api_name="previous_rows")
+
+
+demo.queue()
+demo.launch(debug=True)