• Tested the training result from 08-13, no big difference.
  • Decide to switch to other embedding models.