Sylheti-MBart

Model Overview

This is a fine-tuned facebook/mbart-large-50-many-to-many-mmt model for translating between Bangla and Sylheti, handling word and sentence translations.

Training Info

• Base Model: facebook/mbart-large-50-many-to-many-mmt
• Dataset Entries: 262
• Training Data Points: 499
• Train Size: 449
• Validation Size: 50
• Epochs: 6
• Batch Size: 1
• Learning Rate: 1.5e-05
• Final Training Loss: 9.2842
• Final Validation Loss: 4.042934894561768
• Training Date: 2025-05-01 07:01:29
• Environment: Google Colab (GPU)

How to Use

Load the model from Hugging Face:

• Repository: kamilhussen24/Sylheti-MBart
• Example:
• Input "কি সত্যি নাকি" (Bangla to Sylheti) Output "কিতা হাছা নি"
• Input "ইগু ফুরি নি" (Sylheti to Bangla) Output "এটা মেয়ে নাকি"

Contact

For issues, contact: kamilhussen24