Upgrade gradio

app.py

@@ -1,92 +1,145 @@
 import gradio as gr
 from transformers import pipeline
+from typing import Dict, List, Union
+import logging
+from pydantic import BaseModel, Field
 
-# Define the models
-model1 = pipeline("text-classification", model="albertmartinez/bert-sdg-classification")
-model2 = pipeline("text-classification", model="albertmartinez/bert-multilingual-sdg-classification")
-model3 = pipeline("text-classification", model="albertmartinez/distilbert-multilingual-sdg-classification")
-model4 = pipeline("text-classification", model="albertmartinez/xlm-roberta-large-sdg-classification")
+# Logging configuration
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
+# Constants
+MAX_LENGTH = 512
+TOP_K = 16
+MODELS = {
+    "bert": "albertmartinez/bert-sdg-classification",
+    "bert-multilingual": "albertmartinez/bert-multilingual-sdg-classification",
+    "distilbert-multilingual": "albertmartinez/distilbert-multilingual-sdg-classification",
+    "xlm-roberta-large": "albertmartinez/xlm-roberta-large-sdg-classification"
+}
 
-def classify_text(text, model):
-    result = model(text, top_k=16, truncation=True, max_length=512)
-    return {p["label"]: p["score"] for p in result}
+class ClassificationResult(BaseModel):
+    """Model for classification results."""
+    label: str = Field(..., description="Classification label")
+    score: float = Field(..., description="Classification score")
 
+def load_models() -> Dict[str, pipeline]:
+    """Load all required models."""
+    try:
+        return {
+            name: pipeline("text-classification", model=model_path)
+            for name, model_path in MODELS.items()
+        }
+    except Exception as e:
+        logger.error(f"Error loading models: {str(e)}")
+        raise
 
-def classify_all(text):
-    return [
-        {p["label"]: p["score"] for p in model1(text, top_k=16, truncation=True, max_length=512)},
-        {p["label"]: p["score"] for p in model2(text, top_k=16, truncation=True, max_length=512)},
-        {p["label"]: p["score"] for p in model3(text, top_k=16, truncation=True, max_length=512)},
-        {p["label"]: p["score"] for p in model4(text, top_k=16, truncation=True, max_length=512)}
-    ]
+# Load models
+models = load_models()
 
+def validate_input(text: str) -> bool:
+    """Validate input text."""
+    if not text or not isinstance(text, str):
+        return False
+    if len(text.strip()) == 0:
+        return False
+    return True
 
-ifaceall = gr.Interface(
-    fn=classify_all,
-    inputs=gr.Textbox(lines=2, label="Text", placeholder="Enter text here..."),
-    outputs=[gr.Label(label="bert"), gr.Label(label="bert-multilingual"), gr.Label(label="distilbert-multilingual"),
-             gr.Label(label="xlm-roberta-large")],
-    title="SDG text classification",
-    description="Enter a text and see the text classification result!",
-    flagging_mode="never",
-    api_name="classify_all"
-)
+def classify_text(text: str, model: pipeline) -> Dict[str, float]:
+    """
+    Classify text using the specified model.
+    
+    Args:
+        text: Text to classify
+        model: Classification model
+        
+    Returns:
+        Dictionary with labels and their scores
+    """
+    if not validate_input(text):
+        raise ValueError("Invalid text input")
+    
+    try:
+        result = model(text, top_k=TOP_K, truncation=True, max_length=MAX_LENGTH)
+        return {p["label"]: p["score"] for p in result}
+    except Exception as e:
+        logger.error(f"Error in classification: {str(e)}")
+        raise
 
-# Interface for the first model
-iface1 = gr.Interface(
-    fn=lambda text: classify_text(text, model1),
-    inputs=gr.Textbox(lines=2, label="Text", placeholder="Enter text here..."),
-    outputs=gr.Label(label="Top SDG Predicted"),
-    title="BERT SDG classification",
-    description="Enter a text and see the text classification result!",
-    flagging_mode="never",
-    api_name="classify_bert"
-)
+def classify_all(text: str) -> List[Dict[str, float]]:
+    """
+    Classify text using all available models.
+    
+    Args:
+        text: Text to classify
+        
+    Returns:
+        List of dictionaries with predictions from each model
+    """
+    if not validate_input(text):
+        raise ValueError("Invalid text input")
+    
+    try:
+        return [
+            classify_text(text, model)
+            for model in models.values()
+        ]
+    except Exception as e:
+        logger.error(f"Error in multiple classification: {str(e)}")
+        raise
 
-# Interface for the second model
-iface2 = gr.Interface(
-    fn=lambda text: classify_text(text, model2),
-    inputs=gr.Textbox(lines=2, label="Text", placeholder="Enter text here..."),
-    outputs=gr.Label(label="Top SDG Predicted"),
-    title="BERT multilingual SDG classification",
-    description="Enter a text and see the text classification result!",
-    flagging_mode="never",
-    api_name="classify_bert-multilingual"
-)
+def create_interface(model_name: str, model: pipeline) -> gr.Interface:
+    """Create a Gradio interface for a specific model."""
+    return gr.Interface(
+        fn=lambda text: classify_text(text, model),
+        inputs=gr.Textbox(
+            lines=2,
+            label="Text",
+            placeholder="Enter text here..."
+        ),
+        outputs=gr.Label(label="Top SDG Predicted"),
+        title=f"{model_name.upper()} SDG classification",
+        description="Enter text and see the classification results!",
+        examples=[["To respond to these new challenges PSV has been encouraged to develop strong links with business and enterprises, to cover remote areas and to cater for less affluent students. It analyses the impact of the dual sector universities in Australia, short cycle higher education in Scotland, three sub-sectors of Norway's tertiary education and the emerging non-university higher education in Italy, as well as the vocational education and training in Spain. It discusses the issues of transition, participation and collaboration of different types of post-secondary education and analysis the impact of policy changes."]],
+        flagging_mode="never",
+        api_name=f"classify_{model_name}"
+    )
 
-# Interface for the three model
-iface3 = gr.Interface(
-    fn=lambda text: classify_text(text, model3),
-    inputs=gr.Textbox(lines=2, label="Text", placeholder="Enter text here..."),
-    outputs=gr.Label(label="Top SDG Predicted"),
-    title="DISTILBERT multilingual SDG classification",
-    description="Enter a text and see the text classification result!",
+# Create main interface
+ifaceall = gr.Interface(
+    fn=classify_all,
+    inputs=gr.Textbox(
+        lines=2,
+        label="Text",
+        placeholder="Enter text here..."
+    ),
+    outputs=[
+        gr.Label(label="bert"),
+        gr.Label(label="bert-multilingual"),
+        gr.Label(label="distilbert-multilingual"),
+        gr.Label(label="xlm-roberta-large")
+    ],
+    title="SDG text classification",
+    description="Enter text and see the classification results!",
+    examples=[["To respond to these new challenges PSV has been encouraged to develop strong links with business and enterprises, to cover remote areas and to cater for less affluent students. It analyses the impact of the dual sector universities in Australia, short cycle higher education in Scotland, three sub-sectors of Norway's tertiary education and the emerging non-university higher education in Italy, as well as the vocational education and training in Spain. It discusses the issues of transition, participation and collaboration of different types of post-secondary education and analysis the impact of policy changes."]],
     flagging_mode="never",
-    api_name="classify_distilbert-multilingual"
+    api_name="classify_all_models"
 )
 
-# Interface for the four model
-iface4 = gr.Interface(
-    fn=lambda text: classify_text(text, model4),
-    inputs=gr.Textbox(lines=2, label="Text", placeholder="Enter text here..."),
-    outputs=gr.Label(label="Top SDG Predicted"),
-    title="XLM-ROBERTA-LARGE SDG classification",
-    description="Enter a text and see the text classification result!",
-    flagging_mode="never",
-    api_name="classify_xlm-roberta-large"
-)
+# Create individual interfaces
+interfaces = [ifaceall] + [
+    create_interface(name, model)
+    for name, model in models.items()
+]
 
-with gr.Blocks() as demo:
-    # Combine both interfaces into a tabbed interface
+# Create main interface with tabs
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
     gr.TabbedInterface(
-        interface_list=[ifaceall, iface1, iface2, iface3, iface4],
-        tab_names=["ALL", "bert-sdg-classification", "bert-multilingual-sdg-classification",
-                   "distilbert-multilingual-sdg-classification", "xlm-roberta-large-sdg-classification"],
-        title="Sustainable Development Goals (SDG) Text Classifier App",
-        theme='base'
+        interface_list=interfaces,
+        tab_names=["ALL"] + list(MODELS.keys()),
+        title="Sustainable Development Goals (SDG) Text Classifier App"
     )
 
 if __name__ == "__main__":
-    print(gr.__version__)
+    logger.info(f"Gradio version: {gr.__version__}")
     demo.launch()

requirements.txt

@@ -1,3 +1,7 @@
-transformers
-torch
-gradio
+gradio==5.33.1
+transformers>=4.30.0
+torch>=2.0.0
+numpy>=1.24.0
+pydantic>=2.7.4,<3.0.0
+fastapi==0.115.9
+websockets>=10.0,<12.0