| language: | |
| - ar | |
| The QARiB-Emotion-600k (QE6) model was trained by continuing the training from the last checkpoint of the QE3 model using the emotion tweets dataset (roughly 1.2 million tweets). | |
| The QE6 model trained for an additional 300k steps (600k in total including the QE3 training steps) and around 31 epochs. |