Mixed keras imports causing version conflicts

[iAmGiG]

Problem

Code mixes standalone Keras imports with TensorFlow Keras imports, which can cause version conflicts and unpredictable behavior.

Example in train_og.py:

from keras.models import load_model          # Standalone keras
from tensorflow.keras.models import Model     # TensorFlow keras
from tensorflow.keras.layers import Input, Dense

Files affected:

• anomaly-detection/train_og.py
• anomaly-detection/test.py
• Other training scripts

Fix

Use tensorflow.keras exclusively:

from tensorflow.keras.models import load_model, Model
from tensorflow.keras.layers import Input, Dense

Priority

MODERATE - Can cause model loading failures and compatibility issues

[iAmGiG]

✅ FIXED and TESTED

What was done:

• Standardized all keras imports to use tensorflow.keras
• Files fixed:
  • anomaly-detection/train_autoencoder.py:9
  • anomaly-detection/test_autoencoder.py:6
  • classification/train_classifier.py:13-16
  • classification/test_classifier.py:10

Changes:

• Changed from keras.models import load_model → from tensorflow.keras.models import load_model
• Changed from keras.layers import Dense → from tensorflow.keras.layers import Dense
• Consolidated all imports to use tensorflow.keras namespace

Testing:

• All scripts import successfully
• Classification training: 99.85% accuracy
• Anomaly detection training: completed successfully
• No version conflicts observed

Status: Ready for TensorFlow 2.x environments