Update app.py

app.py

@@ -2,28 +2,104 @@ import gradio as gr
 from transformers import AutoTokenizer, AutoModelForSequenceClassification
 import torch
 
+APP_NAME = "EuroChef"
+
 tokenizer = AutoTokenizer.from_pretrained("BenTouss/mdeberta-eurochef")
 model = AutoModelForSequenceClassification.from_pretrained("BenTouss/mdeberta-eurochef")
 
-def get_labels_table(text):
+device = "cuda" if torch.cuda.is_available() else "cpu"
+model.to(device)
+model.eval()
+
+def predict(text: str, threshold: float = 0.6, top_k: int = 8, only_above: bool = True):
+    text = (text or "").strip()
+    if not text:
+        return "_Colle un message à gauche pour lancer l’analyse._", []
+
     inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=256)
+    inputs = {k: v.to(device) for k, v in inputs.items()}
 
     with torch.no_grad():
         outputs = model(**inputs)
-        probs = torch.sigmoid(outputs.logits)[0]
+        probs = torch.sigmoid(outputs.logits)[0].detach().cpu()
 
-    rows = []
+    items = []
     for idx, prob in enumerate(probs):
-        if prob > 0.6:
-            rows.append([model.config.id2label[idx], float(prob)])
+        score = float(prob)
+        label = model.config.id2label[idx]
+        if (not only_above) or (score >= threshold):
+            items.append((label, score))
+
+    items.sort(key=lambda x: x[1], reverse=True)
+    items = items[: max(1, int(top_k))]
+
+    rows = [[lbl, float(f"{sc:.3f}")] for lbl, sc in items]
+
+    if rows:
+        best_lbl, best_sc = rows[0][0], rows[0][1]
+        summary = f"**Top label :** `{best_lbl}`  •  **score :** `{best_sc}`  \n**Results :** {len(rows)}  •  **threshold :** `{threshold:.2f}`"
+    else:
+        summary = f"_No label (thresold `{threshold:.2f}`)._"
+
+    return summary, rows
+
+
+CSS = """
+#title { margin-bottom: 0.25rem; }
+#subtitle { margin-top: 0; opacity: 0.8; }
+.footer { opacity: 0.7; font-size: 0.85rem; text-align: center; margin-top: 0.75rem; }
+"""
+
+with gr.Blocks(theme=gr.themes.Soft(), css=CSS) as demo:
+    gr.Markdown(f"# 🍳 {APP_NAME}", elem_id="title")
+    gr.Markdown("Customer support message → labels + scores.", elem_id="subtitle")
+
+    with gr.Row():
+        with gr.Column(scale=6):
+            text = gr.Textbox(
+                label="Customer support message",
+                placeholder="Ex: Bonjour, je n’arrive pas à lancer les vidéos…",
+                lines=10,
+            )
+
+            with gr.Row():
+                threshold = gr.Slider(0.0, 1.0, value=0.6, step=0.01, label="Threshold")
+                top_k = gr.Slider(1, 20, value=8, step=1, label="Top-K")
+                only_above = gr.Checkbox(value=True, label="Only ≥ threshold")
+
+            with gr.Row():
+                run = gr.Button("Analyze", variant="primary")
+                clear = gr.ClearButton(value="Clear")
+
+            gr.Examples(
+                examples=[
+                    "Bonjour, je n’arrive pas à lancer les vidéos : écran noir et chargement infini. Je suis Premium mais certaines recettes restent verrouillées…",
+                    "Je veux annuler mon abonnement mais je ne trouve pas où le faire dans l’app.",
+                    "Paiement refusé alors que ma carte fonctionne ailleurs. Pouvez-vous vérifier ?",
+                    "L’application plante dès que je lance une vidéo en Chromecast.",
+                ],
+                inputs=[text],
+                label="Exemples",
+            )
+
+        with gr.Column(scale=6):
+            summary = gr.Markdown(label="Summary")
+            table = gr.Dataframe(
+                headers=["label", "score"],
+                datatype=["str", "number"],
+                label="Predictions",
+                wrap=True,
+                interactive=False,
+                height=320,
+            )
 
-    rows.sort(key=lambda x: x[1], reverse=True)
-    return rows
+    gr.Markdown(f"<div class='footer'>Made with ❤️ by Ben • {APP_NAME}</div>")
 
-demo = gr.Interface(
-    fn=get_labels_table,
-    inputs=gr.Textbox(lines=3, label="Text"),
-    outputs=gr.Dataframe(headers=["label", "score"]),
-)
+    run.click(
+        fn=predict,
+        inputs=[text, threshold, top_k, only_above],
+        outputs=[summary, table],
+    )
+    clear.add([text, summary, table])
 
 demo.launch()