| mBART (a pre-trained model by Facebook) is pre-trained to de-noise multiple languages simultaneously with BART objective. | |
| Checkpoint available in this repository is obtained after fine-tuning facebook/mbart-large-cc25 on 14000 samples from IIT-B Hindi-English parallel corpus. This checkpoint gives decent results for Hindi-english translation. | |