| # Fixed-roberta-base | |
| [roberta-base](https://huggingface.co/roberta-base) but with a resized embedding matrix and an extra dim in the token type embedding matrix for better sharding/partitioning. |
| # Fixed-roberta-base | |
| [roberta-base](https://huggingface.co/roberta-base) but with a resized embedding matrix and an extra dim in the token type embedding matrix for better sharding/partitioning. |