Update app.py

app.py

@@ -1,42 +1,138 @@
 import gradio as gr
-import torch
-import torch.nn as nn
-from transformers import (
-    AutoTokenizer, AutoModel, AutoProcessor, 
-    AutoModelForCausalLM, TrainingArguments, Trainer,
-    DataCollatorForLanguageModeling
-)
-from datasets import Dataset, load_dataset, concatenate_datasets
-import json
 import os
 import requests
-from PIL import Image
-import librosa
-import cv2
-import numpy as np
-from pathlib import Path
+import json
 import logging
 from typing import Dict, List, Optional, Union
 import time
-from huggingface_hub import HfApi, list_datasets_in_collection
 import tempfile
 import shutil
 
+# Imports conditionnels pour éviter les erreurs
+try:
+    import numpy as np
+    NUMPY_AVAILABLE = True
+except ImportError:
+    NUMPY_AVAILABLE = False
+    import array
+    
+try:
+    from pathlib import Path
+    PATHLIB_AVAILABLE = True
+except ImportError:
+    PATHLIB_AVAILABLE = False
+
+try:
+    from huggingface_hub import HfApi
+    HF_HUB_AVAILABLE = True
+except ImportError:
+    HF_HUB_AVAILABLE = False
+    
+try:
+    import numpy as np
+    NUMPY_AVAILABLE = True
+except ImportError:
+    NUMPY_AVAILABLE = False
+    
+try:
+    import torch
+    import torch.nn as nn
+    TORCH_AVAILABLE = True
+except ImportError:
+    TORCH_AVAILABLE = False
+    torch = None
+
+try:
+    from transformers import (
+        AutoTokenizer, AutoModel, AutoProcessor, 
+        AutoModelForCausalLM, TrainingArguments, Trainer,
+        DataCollatorForLanguageModeling
+    )
+    TRANSFORMERS_AVAILABLE = True
+except ImportError:
+    TRANSFORMERS_AVAILABLE = False
+
+try:
+    from datasets import Dataset, load_dataset, concatenate_datasets
+    DATASETS_AVAILABLE = True
+except ImportError:
+    DATASETS_AVAILABLE = False
+
+try:
+    from PIL import Image
+    PIL_AVAILABLE = True
+except ImportError:
+    PIL_AVAILABLE = False
+
+try:
+    import librosa
+    LIBROSA_AVAILABLE = True
+except ImportError:
+    LIBROSA_AVAILABLE = False
+
+try:
+    import cv2
+    CV2_AVAILABLE = True
+except ImportError:
+    CV2_AVAILABLE = False
+
 # Configuration du logging
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
 class MultimodalTrainer:
     def __init__(self):
-        self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+        # Vérification des dépendances
+        self.dependencies_ok = self.check_dependencies()
+        
+        if not TORCH_AVAILABLE:
+            self.device = "cpu"
+            logger.warning("PyTorch non disponible")
+        else:
+            self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+            
         self.current_model = None
         self.current_tokenizer = None
         self.current_processor = None
         self.training_data = []
-        self.hf_api = HfApi()
+        
+        if HF_HUB_AVAILABLE:
+            self.hf_api = HfApi()
+        else:
+            self.hf_api = None
+        
+    def check_dependencies(self):
+        """Vérifie les dépendances installées"""
+        deps = {
+            "PyTorch": TORCH_AVAILABLE,
+            "Transformers": TRANSFORMERS_AVAILABLE,
+            "Datasets": DATASETS_AVAILABLE,
+            "PIL": PIL_AVAILABLE,
+            "Librosa": LIBROSA_AVAILABLE,
+            "OpenCV": CV2_AVAILABLE,
+            "NumPy": NUMPY_AVAILABLE,
+            "HuggingFace Hub": HF_HUB_AVAILABLE
+        }
+        
+        status = "📦 État des dépendances:\n"
+        for name, available in deps.items():
+            status += f"{'✅' if available else '❌'} {name}\n"
+            
+        if not TORCH_AVAILABLE:
+            status += "\n⚠️ PyTorch requis pour l'entraînement!"
+        if not TRANSFORMERS_AVAILABLE:
+            status += "\n⚠️ Transformers requis pour les modèles!"
+            
+        return status
         
     def load_model(self, model_name: str, model_type: str = "causal"):
         """Charge un modèle depuis Hugging Face"""
+        if not TRANSFORMERS_AVAILABLE:
+            return "❌ Transformers non installé!"
+            
+        if not TORCH_AVAILABLE:
+            return "❌ PyTorch non installé!"
+            
         try:
             logger.info(f"Chargement du modèle: {model_name}")
             
@@ -79,12 +175,19 @@ class MultimodalTrainer:
     
     def load_collection_datasets(self, collection_url: str):
         """Charge tous les datasets d'une collection HF"""
+        if not DATASETS_AVAILABLE:
+            return "❌ Datasets non installé!"
+            
         try:
             # Extrait l'ID de la collection depuis l'URL
             collection_id = collection_url.split("/")[-1]
             
-            # Liste les datasets de la collection
-            collection_items = list_datasets_in_collection(collection_id)
+            # Pour l'instant, utilise l'API HF de base
+            try:
+                from huggingface_hub import list_datasets_in_collection
+                collection_items = list_datasets_in_collection(collection_id)
+            except ImportError:
+                return "❌ Fonction collection non disponible, ajoutez manuellement les datasets"
             
             datasets_info = []
             loaded_datasets = []