update app

app.py

@@ -1,115 +1,54 @@
 import gradio as gr
-import spaces
-from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
-from qwen_vl_utils import process_vision_info
-import torch
+from openai import OpenAI
 from PIL import Image
-import subprocess
-from datetime import datetime
 import numpy as np
 import os
+from datetime import datetime
 
-
-# subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
-
-# models = {
-#     "Qwen/Qwen2-VL-7B-Instruct": AutoModelForCausalLM.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True, torch_dtype="auto", _attn_implementation="flash_attention_2").cuda().eval()
+# Initialize OpenAI client
+client = OpenAI()
 
 
-# }
 def array_to_image_path(image_array):
     if image_array is None:
         raise ValueError("No image provided. Please upload an image before submitting.")
-    # Convert numpy array to PIL Image
-    img = Image.fromarray(np.uint8(image_array))
 
-    # Generate a unique filename using timestamp
+    img = Image.fromarray(np.uint8(image_array))
     timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
     filename = f"image_{timestamp}.png"
-
-    # Save the image
     img.save(filename)
-
-    # Get the full path of the saved image
     full_path = os.path.abspath(filename)
-
     return full_path
 
 
-models = {
-    "Qwen/Qwen2-VL-7B-Instruct": Qwen2VLForConditionalGeneration.from_pretrained(
-        "Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True, torch_dtype="auto"
-    )
-    .cpu()
-    .eval()
-}
-
-processors = {
-    "Qwen/Qwen2-VL-7B-Instruct": AutoProcessor.from_pretrained(
-        "Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True
-    )
-}
-
-DESCRIPTION = "WordLift Product Description Generation - [Qwen2-VL-7B Demo](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct)"
-
-kwargs = {}
-kwargs["torch_dtype"] = torch.bfloat16
-
-user_prompt = "<|user|>\n"
-assistant_prompt = "<|assistant|>\n"
-prompt_suffix = "<|end|>\n"
-
-
-@spaces.GPU
-def run_example(image, text_input=None, model_id="Qwen/Qwen2-VL-7B-Instruct"):
+# Function to generate product description using OpenAI API
+def generate_product_description(image, text_input=None):
+    # Convert the image to a path (optional, could directly send the image as a URL if available)
     image_path = array_to_image_path(image)
 
-    print(image_path)
-    model = models[model_id]
-    processor = processors[model_id]
-
-    prompt = f"{user_prompt}<|image_1|>\n{text_input}{prompt_suffix}{assistant_prompt}"
-    image = Image.fromarray(image).convert("RGB")
-    messages = [
-        {
-            "role": "user",
-            "content": [
-                {
-                    "type": "image",
-                    "image": image_path,
-                },
-                {"type": "text", "text": text_input},
-            ],
-        }
-    ]
-
-    # Preparation for inference
-    text = processor.apply_chat_template(
-        messages, tokenize=False, add_generation_prompt=True
-    )
-    image_inputs, video_inputs = process_vision_info(messages)
-    inputs = processor(
-        text=[text],
-        images=image_inputs,
-        videos=video_inputs,
-        padding=True,
-        return_tensors="pt",
-    )
-    inputs = inputs.to("cpu")
-
-    # Inference: Generation of the output
-    generated_ids = model.generate(**inputs, max_new_tokens=1024)
-    generated_ids_trimmed = [
-        out_ids[len(in_ids) :]
-        for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
-    ]
-    output_text = processor.batch_decode(
-        generated_ids_trimmed,
-        skip_special_tokens=True,
-        clean_up_tokenization_spaces=False,
+    # Assuming the image is hosted online, replace the path with the URL.
+    # In practice, you'd need a public URL to share the image with the API.
+    image_url = "https://example.com/" + os.path.basename(image_path)
+
+    # API request
+    completion = client.chat.completions.create(
+        model="gpt-4o",
+        messages=[
+            {
+                "role": "user",
+                "content": [
+                    {"type": "text", "text": text_input or "What's in this image?"},
+                    {
+                        "type": "image_url",
+                        "image_url": {"url": image_url},
+                    },
+                ],
+            }
+        ],
     )
 
-    return output_text[0]
+    # Extract and return the generated message
+    return completion.choices[0].message
 
 
 css = """
@@ -121,23 +60,18 @@ css = """
 """
 
 with gr.Blocks(css=css) as demo:
-    gr.Markdown(DESCRIPTION)
+    gr.Markdown("WordLift Product Description Generation - [GPT-4o-mini Demo]")
     with gr.Tab(label="WordLift Product Description Generation"):
         with gr.Row():
             with gr.Column():
                 input_img = gr.Image(label="Input Picture")
-                model_selector = gr.Dropdown(
-                    choices=list(models.keys()),
-                    label="Model",
-                    value="Qwen/Qwen2-VL-7B-Instruct",
-                )
-                text_input = gr.Textbox(label="Question")
+                text_input = gr.Textbox(label="Additional Instructions (Optional)")
                 submit_btn = gr.Button(value="Submit")
             with gr.Column():
                 output_text = gr.Textbox(label="Output Text")
 
         submit_btn.click(
-            run_example, [input_img, text_input, model_selector], [output_text]
+            generate_product_description, [input_img, text_input], [output_text]
         )
 
 demo.queue(api_open=False)