can know how come model size was too less ?
#3
by
saichand09
- opened
is any c translate or any thing is used for the model ?
It is probably due to the fine-tuning process performed on the Hindi language.
This happens when a subset (i.e. a smaller dataset ) is used while training (fine-tuning on model), resulting in a smaller weight for the trained model deployed for this specific task. Also the vocabulary affects results, hence it would not perform well on different languages beside Hindi (it will probably use only Hindi characters, indeed). I hope it helps.
thanks very much