change requirements and model

app.py

@@ -9,6 +9,7 @@ from transformers import Wav2Vec2Processor, Wav2Vec2Model, AutoTokenizer, AutoMo
 from torchvision import models
 import tempfile
 import os
+from huggingface_hub import hf_hub_download
 
 # Configuration
 DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
@@ -100,15 +101,28 @@ class AVVideoModel(nn.Module):
         return fused_logits, a_logits, t_logits, v_logits
 
 # Load model
+
 model = AVVideoModel(num_classes=len(LABELS)).to(DEVICE)
 
-# Load trained weights (you'll need to upload this)
-if os.path.exists("model_weights.pth"):
-    model.load_state_dict(torch.load("model_weights.pth", map_location=DEVICE))
+# Download model from Hugging Face Model Hub
+try:
+    model_path = hf_hub_download(
+        repo_id="ApurvaKondekar/emotion_model",  # CHANGE THIS
+        filename="model_weights.pth"  # YOUR FILE NAME
+    )
+    model.load_state_dict(torch.load(model_path, map_location=DEVICE))
     model.eval()
-    print("✅ Model loaded successfully")
-else:
-    print("⚠️ No model weights found. Using untrained model for demo.")
+    print("✅ Model loaded from Hugging Face")
+except Exception as e:
+    print(f"❌ Failed to load model: {e}")
+    # Alternative: Create a dummy model for testing
+    # Load trained weights (you'll need to upload this)
+    if os.path.exists("model_weights.pth"):
+        model.load_state_dict(torch.load("model_weights.pth", map_location=DEVICE))
+        model.eval()
+        print("✅ Model loaded successfully")
+    else:
+        print("⚠️ No model weights found. Using untrained model for demo.")
 
 def extract_video_frames(video_path, max_frames=8, resize=(224, 224)):
     """Extract frames from video file"""

requirements.txt

@@ -6,3 +6,4 @@ opencv-python-headless
 numpy
 soundfile
 accelerate
+huggingface_hub