---
language: ti
widget:
- text: "ዓቕሚ ደቀንስትዮ [MASK] ብግብሪ ተራእዩ"
---

# RoBERTa Pretrained for Tigrinya Language

We pretrain a RoBERTa Base model on a relatively small dataset for Tigrinya (34M tokens) for 18 epochs.

Contained in this card is a PyTorch model exported from the original model that was trained on TPU v3.8 with Flax.


## Hyperparameters

The hyperparameters corresponding to model sizes mentioned above are as follows:

| Model Size | L  | AH | HS  | FFN  | P    |
|------------|----|----|-----|------|------|
| BASE       | 12 | 12 | 768 | 3072 | 125M |

(L = number of layers; AH = number of attention heads; HS = hidden size; FFN = feedforward network dimension; P = number of parameters.)