| license: mit | |
| library_name: keras | |
| pipeline_tag: fill-mask | |
| tags: | |
| - keras | |
| - kerasformers | |
| - tf | |
| - jax | |
| - pytorch | |
| - backend-agnostic | |
| - deberta_v2 | |
| - fill-mask | |
| base_model: | |
| - microsoft/deberta-v2-xlarge | |
| # DeBERTa-v2 (deberta_v2_xlarge) | |
| A pure [Keras 3](https://keras.io) port of **DeBERTa-v2**, converted from [`microsoft/deberta-v2-xlarge`](https://huggingface.co/microsoft/deberta-v2-xlarge) (Hugging Face Transformers). | |
| The weights are **backend-agnostic**: the same checkpoint loads and runs identically under the JAX, TensorFlow, or PyTorch Keras backend (`KERAS_BACKEND=jax|tensorflow|torch`), via [kerasformers](https://github.com/IMvision12/KerasFormers). | |
| ## Usage | |
| ```python | |
| from kerasformers.models.deberta_v2 import DebertaV2Model | |
| model = DebertaV2Model.from_weights("kerasformers/deberta_v2_xlarge") | |
| ``` | |