metadata
datasets:
- ameykaran/mizo-text-corpus
- ameykaran/english-text-corpus
- ameykaran/hindi-text-corpus
language:
- hi
- en
metrics:
- perplexity
base_model:
- deepseek-ai/DeepSeek-V3
library_name: transformers
tags:
- mizo
- english
- hindi
- multilingual
- indic
DilLeiX Model
An indic-multilingual small language model (~150M parameters) capable of understanding English, Hindi and Mizo. It is based on the DeepSeek-V3 architecture and uses a byte-level BPE tokeniser.