Update app.py

app.py

@@ -1,17 +1,15 @@
 import gradio as gr
-from transformers import AutoModelForVision2Seq, AutoProcessor, AutoModelForVision2Seq, BitsAndBytesConfig
+from transformers import AutoModelForVision2Seq, AutoProcessor, BitsAndBytesConfig
 import torch
 
+
+model_id = "HuggingFaceM4/idefics2-8b"
+
 quantization_config = BitsAndBytesConfig(
     load_in_4bit=True, bnb_4bit_quant_type="nf4", bnb_4bit_use_double_quant=True, bnb_4bit_compute_dtype=torch.float16
 )
-
-processor = AutoProcessor.from_pretrained("HuggingFaceM4/idefics2-8b")
-model = AutoModelForVision2Seq.from_pretrained(
-    "HuggingFaceM4/idefics2-8b",
-    torch_dtype=torch.float16,
-    quantization_config=quantization_config,
-)
+processor = AutoProcessor.from_pretrained(model_id)
+model = AutoModelForVision2Seq.from_pretrained(model_id, torch_dtype=torch.float16, quantization_config=quantization_config)
 
 
 def respond(multimodal_input):
@@ -31,4 +29,14 @@ def respond(multimodal_input):
     return generated_text
 
 
-gr.Interface(respond, inputs=[gr.MultimodalTextbox(file_types=["image"], show_label=False)], outputs="text").launch()
+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()