Upload 2 files

app.py

@@ -0,0 +1,81 @@
+import gradio as gr
+from transformers import AutoModelForSequenceClassification, AutoTokenizer
+import torch
+import os
+
+# --- Configuration ---
+MODEL_PATH = "ag_news_distilbert_finetuned" # This is the folder saved in your ZIP file
+TARGET_NAMES = ["World", "Sports", "Business", "Sci/Tech"]
+
+# Check if the model files are present (critical for local/Spaces deployment)
+if not os.path.exists(MODEL_PATH):
+    print(f"Error: Model directory '{MODEL_PATH}' not found. Ensure you have extracted the ZIP file.")
+    # Exit or raise error if running locally
+    # If running on HF Spaces, the folder structure must match.
+
+# --- 1. Load Model and Tokenizer (The Persistence Step) ---
+try:
+    # Use the local model path to load the fine-tuned weights
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
+    model = AutoModelForSequenceClassification.from_pretrained(MODEL_PATH)
+    model.eval() # Set the model to evaluation mode
+    print("Model and Tokenizer loaded successfully from local directory.")
+except Exception as e:
+    print(f"Failed to load model from path: {e}")
+    # Fallback or exit logic if needed
+
+
+# --- 2. Define the Prediction Function ---
+def classify_news(text):
+    """Takes input text, tokenizes it, and returns the predicted class name and confidence scores."""
+    if not text or len(text.strip()) < 5:
+        return "Please enter a longer news snippet or headline.", {}
+    
+    # a. Tokenize and prepare input
+    inputs = tokenizer(text, return_tensors="pt", padding=True, truncation=True)
+    
+    # b. Run inference (crucial to use torch.no_grad() for efficient inference)
+    with torch.no_grad():
+        outputs = model(**inputs)
+        logits = outputs.logits
+        probabilities = torch.softmax(logits, dim=1)[0].tolist()
+        
+    # c. Map probabilities to labels
+    confidence = {
+        TARGET_NAMES[i]: prob for i, prob in enumerate(probabilities)
+    }
+    
+    # d. Return the results
+    return confidence
+
+# --- 3. Create the Gradio Interface ---
+iface = gr.Interface(
+    fn=classify_news,
+    inputs=gr.Textbox(
+        lines=5, 
+        label="Input News Headline or Snippet",
+        placeholder="Example: Russia and Canada discuss gas pipeline auction..."
+    ),
+    outputs=gr.Label(
+        num_top_classes=4, # Display all 4 classes and their scores
+        label="Predicted Category and Confidence"
+    ), 
+    title="AG News Classifier: Fine-Tuned DistilBERT",
+    description=(
+        "Enter a news headline or short article. The model will predict its category "
+        "(World, Sports, Business, or Sci/Tech) and show the confidence scores for all categories."
+    ),
+    # Add examples from your analysis to show it handles tricky inputs
+    examples=[
+        ["The global stock market rallied today after the central bank cut interest rates."], # Business
+        ["New research shows quantum entanglement may enable faster computing."], # Sci/Tech
+        ["Manchester United defeats Liverpool in a stunning Premier League match."], # Sports
+        ["The President's cabinet held an emergency summit on trade negotiations."], # World
+        ["AT&T Wireless ships mobile IM gadget US mobile network operator AT&T Wireless today launched Ogo, its first non-voice messaging device, pitche..."] # Misclassified Example
+    ],
+    allow_flagging="never", # Optional: prevents user from submitting feedback
+)
+
+# Launch the application
+if __name__ == "__main__":
+    iface.launch()

requirements.txt

@@ -0,0 +1,3 @@
+transformers
+torch
+gradio