Upload app.py with huggingface_hub

app.py

@@ -51,15 +51,11 @@ ID2LABEL = {0: "fake", 1: "real"}
 def predict_fake_news(headline: str, source: str):
     headline = headline.strip()
     if not headline:
-        return {"label": "Please enter a headline.", "confidence": None, "probabilities": {"real": 0, "fake": 0}}
+        return "Please enter a headline.", None
 
     # Shortcut: if source is known authentic, return “real”
     if source in AUTHENTIC_SOURCES:
-        return {
-            "label": "real",
-            "confidence": 0.99,
-            "probabilities": {"real": 0.99, "fake": 0.01}
-        }
+        return "real", 0.99
 
     # Otherwise, run transformer inference
     encoding = tokenizer(
@@ -74,22 +70,17 @@ def predict_fake_news(headline: str, source: str):
         outputs = model(**encoding)
         logits = outputs.logits.squeeze()
         probs = F.softmax(logits, dim=-1)
-        fake_conf, real_conf = float(probs[0].cpu().item()), float(probs[1].cpu().item())
         confidence, pred_id = torch.max(probs, dim=-1)
         label = ID2LABEL[int(pred_id)]
 
-    return {
-        "label": label,
-        "confidence": float(confidence.cpu().item()),
-        "probabilities": {"real": real_conf, "fake": fake_conf}
-    }
+    return label, float(confidence.cpu().item())
 
 def load_example(example):
     # Return headline, default source, and clear outputs
-    return example, ALL_SOURCES[0], "", "", {"real": 0, "fake": 0}
+    return example, ALL_SOURCES[0], "", None
 
 def reset_inputs():
-    return "", ALL_SOURCES[0], "", "", {"real": 0, "fake": 0}
+    return "", ALL_SOURCES[0], "", None
 
 with gr.Blocks() as demo:
     gr.Markdown("<h1 style='text-align: center'>📰 Fake News Detector</h1>")
@@ -118,30 +109,21 @@ with gr.Blocks() as demo:
             with gr.Column(scale=1):
                 label_output = gr.Textbox(label="Predicted Label", interactive=False)
                 confidence_output = gr.Textbox(label="Confidence (0–1)", interactive=False)
-                prob_bar = gr.BarPlot(
-                    value="probabilities",
-                    label="Probabilities",
-                    x="class",
-                    y="prob",
-                    color="class",
-                    show_legend=False,
-                    title="Confidence Distribution"
-                )
 
         example_dropdown.change(
             fn=load_example,
             inputs=example_dropdown,
-            outputs=[headline_input, source_dropdown, label_output, confidence_output, prob_bar]
+            outputs=[headline_input, source_dropdown, label_output, confidence_output]
         )
         predict_button.click(
             fn=predict_fake_news,
             inputs=[headline_input, source_dropdown],
-            outputs=[label_output, confidence_output, prob_bar]
+            outputs=[label_output, confidence_output]
         )
         reset_button.click(
             fn=reset_inputs,
             inputs=None,
-            outputs=[headline_input, source_dropdown, label_output, confidence_output, prob_bar]
+            outputs=[headline_input, source_dropdown, label_output, confidence_output]
         )
 
     with gr.Tab("About"):
@@ -163,10 +145,9 @@ with gr.Blocks() as demo:
 
             **Features:**
             - **Example Headlines:** Quickly load sample headlines for testing.  
-            - **Probability Bar:** Visualize model confidence for both classes.  
             - **Source Shortcut:** Automatically mark known authentic sources as real.  
 
-            **Creator:** Your Name or Team  
+            **Creator:** Blessing 
             **Model:** Fine-tuned DistilBERT for Fake News Detection  
             **License:** MIT License
             """