Update app.py

app.py

@@ -1,15 +1,13 @@
 import gradio as gr
 from transformers import AutoProcessor, AutoModelForCausalLM
 import spaces
+
 import io
-import base64  # Adicionando a biblioteca base64 para decodificação
 from PIL import Image
+import base64  # Para decodificar imagens Base64
 import subprocess
-
-# Instalando a dependência flash-attn se necessário
 subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
-
-# Carregando o modelo e o processador
+#
 model_id = 'J-LAB/Florence-vl3'
 model = AutoModelForCausalLM.from_pretrained(model_id, trust_remote_code=True).to("cuda").eval()
 processor = AutoProcessor.from_pretrained(model_id, trust_remote_code=True)
@@ -35,35 +33,28 @@ def run_example(task_prompt, image):
     )
     return parsed_answer
 
-# Função para processar imagens, agora suportando Base64
 def process_image(image, task_prompt):
-    # Verifica se a imagem é uma string base64
-    if isinstance(image, str) and image.startswith("data:image"):
-        # Extraindo a parte base64 da string
-        base64_image = image.split(",")[1]
-        # Decodificando a imagem base64
-        image = Image.open(io.BytesIO(base64.b64decode(base64_image)))
-    elif isinstance(image, bytes):
-        image = Image.open(io.BytesIO(image))
-    else:
-        image = Image.fromarray(image)  # Convertendo um array NumPy para imagem PIL, se aplicável
-
-    # Mapeando os prompts de tarefas
+    if isinstance(image, str):
+        if image.startswith('data:image/png;base64,'):
+            # Decodifica a imagem Base64
+            image_data = base64.b64decode(image.split(',')[1])
+            image = Image.open(io.BytesIO(image_data))
+    
+    image = Image.fromarray(image)  # Convert NumPy array to PIL Image
     if task_prompt == 'Product Caption':
         task_prompt = '<MORE_DETAILED_CAPTION>'
     elif task_prompt == 'OCR':
         task_prompt = '<OCR>'
     
-    # Chamando o exemplo com a imagem processada e o prompt da tarefa
     results = run_example(task_prompt, image)
     
-    # Extraindo o texto gerado a partir dos resultados
+    # Remove the key and get the text value
     if results and task_prompt in results:
         output_text = results[task_prompt]
     else:
         output_text = ""
 
-    # Convertendo quebras de linha para quebras de linha HTML
+    # Convert newline characters to HTML line breaks
     output_text = output_text.replace("\n\n", "<br><br>").replace("\n", "<br>")
 
     return output_text
@@ -92,47 +83,41 @@ document.querySelector('button').addEventListener('click', function() {
 });
 """
 
-single_task_list = ['Product Caption', 'OCR']
+single_task_list =[
+    'Product Caption', 'OCR'
+]
 
 with gr.Blocks(css=css) as demo:
     gr.Markdown(DESCRIPTION)
     with gr.Tab(label="Product Image Select"):
         with gr.Row():
             with gr.Column():
-                input_img = gr.Image(label="Input Picture", source="upload", type="pil")  # Suporte a PIL images
+                input_img = gr.Image(label="Input Picture")
                 task_prompt = gr.Dropdown(choices=single_task_list, label="Task Prompt", value="Product Caption")
                 submit_btn = gr.Button(value="Submit")
             with gr.Column():
                 output_text = gr.HTML(label="Output Text", elem_id="output")
         
-        gr.Markdown(""" 
+        gr.Markdown("""
         ## How to use via API
         To use this model via API, you can follow the example code below:
 
-        ```python
-        import base64
-        from PIL import Image
-        import io
-        import requests
-
-        # Converting image to base64
-        image_path = 'path_to_image.png'
-        with open(image_path, 'rb') as image_file:
-            image_base64 = base64.b64encode(image_file.read()).decode('utf-8')
         
-        # Preparing the payload
-        payload = {
-            "image": f"data:image/png;base64,{image_base64}",
-            "task_prompt": "Product Caption"
-        }
-
-        response = requests.post("http://your-space-url-here", json=payload)
-        print(response.json())
-        ``` 
+python
+        !pip install gradio_client
+        from gradio_client import Client, handle_file
+
+        client = Client("J-LAB/Fluxi-IA")
+        result = client.predict(
+            image=handle_file('https://raw.githubusercontent.com/gradio-app/gradio/main/test/test_files/bus.png'),
+            api_name="/process_image"
+        )
+        print(result)
+
         """)
         
         submit_btn.click(process_image, [input_img, task_prompt], [output_text])
 
     demo.load(lambda: None, inputs=None, outputs=None, js=js)
 
-demo.launch(debug=True)
+demo.launch(debug=True)