Upload folder using huggingface_hub

exemples/demo_batch_hf.py

@@ -1,24 +1,31 @@
 #!/usr/bin/env python3
 """
-📦 Prédiction BATCH via API Hugging Face
+📦 Prédiction BATCH via API Hugging Face (Gradio Client)
 
 Usage: python demo_batch_hf.py
 - Utilise par défaut les CSV d'exemple du dossier
-- Envoie les 3 fichiers à la Space HF
+- Envoie les 3 fichiers à la Space HF via Gradio Client
 - Sauvegarde un CSV de résultats
 
-Option: définir HF_API_URL pour surcharger l'URL par défaut.
+Prérequis: pip install gradio_client
 """
 
 import os
+import sys
 import pandas as pd
-import requests
 from datetime import datetime
 
+try:
+    from gradio_client import Client, handle_file
+except ImportError:
+    print("❌ gradio_client non installé. Installez-le avec:")
+    print("   pip install gradio_client")
+    sys.exit(1)
+
 API_URL = os.getenv("HF_API_URL", "https://asi-engineer-oc-p5.hf.space")
 
 print("╔══════════════════════════════════════════════════════════╗")
-print("║  📦 Prédiction BATCH - API Hugging Face                 ║")
+print("║  📦 Prédiction BATCH - API Hugging Face (Gradio)        ║")
 print("╚══════════════════════════════════════════════════════════╝\n")
 print(f"🌐 API: {API_URL}\n")
 
@@ -32,88 +39,71 @@ sirh_path = os.path.join(script_dir, "02_predict_batch_sirh.csv")
 for path in [sondage_path, eval_path, sirh_path]:
     if not os.path.exists(path):
         print(f"❌ Fichier introuvable: {path}")
-        raise SystemExit(1)
+        sys.exit(1)
 
 print("✅ Fichiers d'exemple détectés:")
 print(f"   - {os.path.basename(sondage_path)}")
 print(f"   - {os.path.basename(eval_path)}")
 print(f"   - {os.path.basename(sirh_path)}\n")
 
-print("⏳ Envoi des fichiers à l'API HF...")
-files = {
-    "sondage_file": open(sondage_path, "rb"),
-    "eval_file": open(eval_path, "rb"),
-    "sirh_file": open(sirh_path, "rb"),
-}
-headers = {}
-api_key = os.getenv("HF_API_KEY")
-if api_key:
-    headers["X-API-Key"] = api_key
+print("⏳ Connexion à l'API Gradio...")
+try:
+    client = Client(API_URL)
+    print("✅ Connecté à l'API Gradio\n")
+except Exception as e:
+    print(f"❌ Impossible de se connecter: {e}")
+    sys.exit(1)
 
+print("⏳ Envoi des fichiers pour prédiction batch...")
 try:
-    # 1) Tente FastAPI (si exposé)
-    r = requests.post(
-        f"{API_URL}/predict/batch", files=files, headers=headers, timeout=90
+    result = client.predict(
+        sondage_path=handle_file(sondage_path),
+        eval_path=handle_file(eval_path),
+        sirh_path=handle_file(sirh_path),
+        api_name="/predict_batch",
     )
-    if r.status_code == 404:
-        # 2) Fallback: endpoint Gradio API
-        print(
-            "\nℹ️ Endpoint FastAPI indisponible, tentative via Gradio API (/api/predict_batch)..."
-        )
-        r = requests.post(
-            f"{API_URL}/api/predict_batch", files=files, headers=headers, timeout=90
-        )
-        if r.status_code == 404:
-            print(
-                "\n❌ Endpoint HF introuvable (/predict/batch et /api/predict_batch)."
-            )
-            print(
-                "   Vérifiez que la Space expose l'API FastAPI ou l'onglet Batch Gradio."
-            )
-            print("   Sinon, utilisez l'API locale (lancer_api.sh).")
-            raise SystemExit(1)
-    r.raise_for_status()
-    result = r.json()
-
-    # Construire le CSV de sortie
-    predictions_data = []
-    for pred in result.get("predictions", []):
-        predictions_data.append(
-            {
-                "employee_id": pred.get("employee_id"),
-                "prediction": (
-                    "VA PARTIR" if pred.get("prediction") == 1 else "VA RESTER"
-                ),
-                "prediction_code": pred.get("prediction"),
-                "risk_level": pred.get("risk_level"),
-                "probability_stay": f"{pred.get('probability_stay', 0):.2%}",
-                "probability_leave": f"{pred.get('probability_leave', 0):.2%}",
-            }
-        )
-
-    df = pd.DataFrame(predictions_data)
-    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
-    output_path = os.path.join(script_dir, f"predictions_batch_hf_{timestamp}.csv")
-    df.to_csv(output_path, index=False, encoding="utf-8-sig")
-
-    # Affichage
-    summary = result.get("summary", {})
-    print("\n" + "═" * 60)
-    print("                    📊 RÉSULTAT (HF)")
-    print("═" * 60)
-    print(
-        f"\n✅ Traités: {result.get('total_employees')} | RESTER: {summary.get('total_stay')} | PARTIR: {summary.get('total_leave')}"
-    )
-    print(
-        f"🔴 High: {summary.get('high_risk_count')}  🟡 Medium: {summary.get('medium_risk_count')}  🟢 Low: {summary.get('low_risk_count')}\n"
+except Exception as e:
+    print(f"❌ Erreur lors de la prédiction: {e}")
+    sys.exit(1)
+
+# Vérifier si erreur dans le résultat
+if isinstance(result, dict) and "error" in result:
+    print(f"\n❌ Erreur API: {result.get('error')}")
+    print(f"   Message: {result.get('message')}")
+    sys.exit(1)
+
+# Construire le CSV de sortie
+predictions_data = []
+for pred in result.get("predictions", []):
+    predictions_data.append(
+        {
+            "employee_id": pred.get("employee_id"),
+            "prediction": "VA PARTIR" if pred.get("prediction") == 1 else "VA RESTER",
+            "prediction_code": pred.get("prediction"),
+            "risk_level": pred.get("risk_level"),
+            "probability_stay": f"{pred.get('probability_stay', 0):.2%}",
+            "probability_leave": f"{pred.get('probability_leave', 0):.2%}",
+        }
     )
-    print("📄 Aperçu:")
-    print(df.head(5).to_string(index=False))
-    print(f"\n💾 Sauvegardé: {output_path}")
-
-finally:
-    for f in files.values():
-        try:
-            f.close()
-        except Exception:
-            pass
+
+df = pd.DataFrame(predictions_data)
+timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+output_path = os.path.join(script_dir, f"predictions_batch_hf_{timestamp}.csv")
+df.to_csv(output_path, index=False, encoding="utf-8-sig")
+
+# Affichage résumé
+summary = result.get("summary", {})
+total = result.get("total_employees", len(predictions_data))
+
+print("\n" + "=" * 50)
+print("📊 RÉSULTATS DE LA PRÉDICTION BATCH")
+print("=" * 50)
+print(f"\n👥 Total employés analysés: {total}")
+print(f"✅ Vont rester:  {summary.get('total_stay', 'N/A')}")
+print(f"❌ Vont partir:  {summary.get('total_leave', 'N/A')}")
+print(f"\n🔴 Risque élevé:  {summary.get('high_risk_count', 'N/A')}")
+print(f"🟠 Risque moyen:  {summary.get('medium_risk_count', 'N/A')}")
+print(f"🟢 Risque faible: {summary.get('low_risk_count', 'N/A')}")
+
+print(f"\n💾 Résultats sauvegardés: {os.path.basename(output_path)}")
+print("\n✅ Prédiction batch terminée avec succès!")