Upload backend/hue_portal/hue-portal-backendDocker/set_14b_model.py with huggingface_hub

backend/hue_portal/hue-portal-backendDocker/set_14b_model.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python3
+"""
+Script to set up Qwen 2.5 14B with 4-bit quantization on Hugging Face Spaces.
+14B model is smaller and faster than 32B, good balance between quality and speed.
+"""
+
+import os
+from huggingface_hub import HfApi
+from huggingface_hub.utils import HfFolder
+
+def get_hf_token():
+    """Get Hugging Face token from cache."""
+    token = HfFolder.get_token()
+    if not token:
+        print("❌ No Hugging Face token found!")
+        print("💡 Run: huggingface-cli login")
+        return None
+    return token
+
+def set_secrets(api, repo_id, secrets):
+    """Set secrets on Hugging Face Space."""
+    try:
+        # Delete existing secrets first to avoid collisions
+        existing_secrets = api.get_space_variables(repo_id=repo_id, token=api.token)
+        existing_secret_names = [s.key for s in existing_secrets if hasattr(s, 'key')]
+        
+        for secret_name in secrets.keys():
+            if secret_name in existing_secret_names:
+                try:
+                    api.delete_space_variable(repo_id=repo_id, key=secret_name, token=api.token)
+                    print(f"🗑️  Deleted existing secret: {secret_name}")
+                except Exception as e:
+                    print(f"⚠️  Could not delete {secret_name}: {e}")
+        
+        # Add new secrets
+        for key, value in secrets.items():
+            api.add_space_secret(repo_id=repo_id, key=key, value=value, token=api.token)
+            print(f"✅ Set secret: {key}")
+        
+        return True
+    except Exception as e:
+        print(f"❌ Error setting secrets: {e}")
+        return False
+
+def main():
+    repo_id = "davidtran999/hue-portal-backend"
+    
+    print("🚀 Setting up Qwen 2.5 14B with 4-bit quantization on HF Space")
+    print(f"📦 Repository: {repo_id}\n")
+    
+    # Get token
+    token = get_hf_token()
+    if not token:
+        return
+    
+    api = HfApi(token=token)
+    
+    # Configuration for 4-bit Qwen 2.5 14B
+    secrets = {
+        "LLM_PROVIDER": "local",
+        "LOCAL_MODEL_PATH": "Qwen/Qwen2.5-14B-Instruct",
+        "LOCAL_MODEL_DEVICE": "cuda",
+        "LOCAL_MODEL_4BIT": "true"
+    }
+    
+    print("📋 Configuration:")
+    for key, value in secrets.items():
+        print(f"   {key}={value}")
+    print()
+    
+    # Set secrets
+    if set_secrets(api, repo_id, secrets):
+        print("\n✅ Successfully configured 4-bit Qwen 2.5 14B!")
+        print("\n📊 Expected Memory Usage:")
+        print("   - VRAM: ~6-8GB (với 4-bit)")
+        print("   - RAM: ~12GB+")
+        print("   - Quality: ⭐⭐⭐⭐ (very good)")
+        print("   - Speed: ⚡⚡⚡ (faster than 32B)")
+        print("\n⚠️  Note: Model will be downloaded on first run (~30GB)")
+        print("   First load may take 10-20 minutes.")
+        print("\n🔄 Rebuild your HF Space to apply changes!")
+        print("   Or wait for auto-rebuild if enabled.")
+    else:
+        print("\n❌ Failed to set secrets")
+
+if __name__ == "__main__":
+    main()
+
+
+