odia-gemma-2b-base (Pre-trained)

Odia-Gemma-2B-Base is a pre-trained Odia large language model with 2 billion parameters, and it is based on Google/Gemma 2B. The model is pre-trained on the Culturex-Odia dataset, a filtered version of the original CulturaX dataset for Odia text. The training dataset contains 49 million tokens. The CulturaX-Odia dataset is sourced from mc4 and four distinct OSCAR corpora.

For more details about the model, data, training procedure, and evaluations, go through the blog post.

Model Description

• Model type: A 2B pre-trained decoder-only model
• Primary Language(s): Odia and English
• License: Gemma Terms of Use

NOTE

This is not an instruction-tuned model, so it may not be able to follow human instructions without using one/few-shot learning or instruction fine-tuning. The model has no moderation mechanisms and may generate harmful or inappropriate responses. It is recommended to first fine-tune it on the task(s) you are interested in.

Citation Information

If you find this model useful, please consider giving 👏 and citing:

@misc{odia-gemma-2b-base,
  author = {Sambit Sekhar and Shantipriya Parida and Debasish Dhal and Guneet Singh Kohli},
  title = {OdiaGenAI Introduces Gemma 2B Pre-Trained LLM Catered to Odia Speakers},
  year = {2024},
  publisher = {Hugging Face},
  journal = {Hugging Face repository},
  howpublished = {\url{https://huggingface.co/OdiaGenAI}},
}

Contributions

• Sambit Sekhar
• Shantipriya Parida
• Debasish Dhal
• Guneet Singh Kohli