nilayshenai/BART-English-to-Bhojpuri-Alpha2

English to Bhojpuri Translation Model Alpha2

The Alpha2 model is a fine-tuned translation model derived from facebook/mbart-large-50-many-to-many-mmt, specifically trained to translate English text into Bhojpuri. It builds on the Alpha1 version, with improvements from training over 4 epochs on a custom parallel dataset.

Space

https://huggingface.co/spaces/nilayshenai/English-to-Bhojpuri-Translator

Updates from Alpha1

• Trained for 4 epochs for better generalization.
• Improved translation fluency and accuracy for Bhojpuri.

Contents

• config.json – Model configuration.
• generation_config.json – Generation parameters (e.g., max length, beam search).
• model.safetensors – Fine-tuned model weights.
• sentencepiece.bpe.model – Tokenizer vocabulary (SentencePiece model).
• special_tokens_map.json – Mapping of special tokens (e.g., BOS, EOS).
• tokenizer.json – Full tokenizer JSON.
• tokenizer_config.json – Tokenizer configuration.

---

license: mit