Update app.py

app.py

@@ -1,84 +1,43 @@
 import gradio as gr
-import spaces
+import torch
 from transformers import Qwen2VLForConditionalGeneration, AutoProcessor
 from qwen_vl_utils import process_vision_info
-import torch
 from PIL import Image
-import subprocess
-from datetime import datetime
 import numpy as np
+from datetime import datetime
 import os
 
-
-# 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()
-
-# }
 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
     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").cuda().eval()
-
-}
-
-processors = {
-    "Qwen/Qwen2-VL-7B-Instruct": AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True)
-}
-
-DESCRIPTION = "[Qwen2-VL-7B Demo](https://huggingface.co/Qwen/Qwen2-VL-7B-Instruct)"
+    return os.path.abspath(filename)
 
-kwargs = {}
-kwargs['torch_dtype'] = torch.bfloat16
+model = Qwen2VLForConditionalGeneration.from_pretrained(
+    "Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True, torch_dtype="auto"
+).cuda().eval()
 
-user_prompt = '<|user|>\n'
-assistant_prompt = '<|assistant|>\n'
-prompt_suffix = "<|end|>\n"
+processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-7B-Instruct", trust_remote_code=True)
 
-@spaces.GPU
-def run_example(image, text_input=None, model_id="Qwen/Qwen2-VL-7B-Instruct"):
+@gr.GPU
+def analyze_iol_report(image):
     image_path = array_to_image_path(image)
+    prompt = "Extract all ophthalmic measurements including AL, K1, K2, ACD, LT, WTW, and CCT. Calculate IOL power based on formulas like Barrett Universal II, Cooke K6, EVO, Hill-RBF, Hoffer QST, Kane, and PEARL GDS."
     
-    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},
+                {"type": "image", "image": image_path},
+                {"type": "text", "text": prompt},
             ],
         }
     ]
     
-    # Preparation for inference
-    text = processor.apply_chat_template(
-        messages, tokenize=False, add_generation_prompt=True
-    )
+    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
     image_inputs, video_inputs = process_vision_info(messages)
     inputs = processor(
         text=[text],
@@ -88,11 +47,9 @@ def run_example(image, text_input=None, model_id="Qwen/Qwen2-VL-7B-Instruct"):
         return_tensors="pt",
     )
     inputs = inputs.to("cuda")
-    
-    # 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)
+        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
@@ -109,18 +66,12 @@ css = """
 """
 
 with gr.Blocks(css=css) as demo:
-    gr.Markdown(DESCRIPTION)
-    with gr.Tab(label="IOL Analyzer"):
+    gr.Markdown("Fixed Model - Qwen2-VL-7B for IOL Report Analysis")
+    with gr.Tab(label="IOL Analysis"):
         with gr.Row():
-            with gr.Column():
-                input_img = gr.Image(label="Input Picture")
-                model_selector = "Qwen/Qwen2-VL-7B-Instruct"
-                text_input = "Please analyze the uploaded IOL report image. Extract all available ophthalmic measurements, including Axial Length (AL), Keratometry Readings (K1, K2), Anterior Chamber Depth (ACD), Lens Thickness (LT), White-to-White Distance (WTW), and Central Corneal Thickness (CCT). Using these measurements, provide IOL power calculations based on Barrett Universal II, Cooke K6, EVO, Hill-RBF, Hoffer QST, Kane, PEARL GDS, and any other relevant formulas. Summarize all calculated values with any additional observations relevant to IOL selection."
-                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])
+            input_img = gr.Image(label="Upload IOL Report Image")
+            submit_btn = gr.Button(value="Analyze Report")
+        output = gr.Textbox(label="Analysis Result", elem_id="output")
+        submit_btn.click(analyze_iol_report, inputs=[input_img], outputs=[output])
 
-demo.queue(api_open=False)
-demo.launch(debug=True)
+demo.launch()