Update README.md

README.md

@@ -1,3 +1,47 @@
----
-license: cc-by-4.0
----
+---
+license: cc-by-4.0
+tags:
+  - translation
+  - onnx
+  - marian
+  - android
+language:
+  - de
+  - ar
+---
+
+# opus-mt-de-ar (ONNX)
+
+ONNX export of [Helsinki-NLP/opus-mt-de-ar](https://huggingface.co/Helsinki-NLP/opus-mt-de-ar) for on-device inference on Android.
+
+## Files
+
+| File | Description |
+|---|---|
+| `encoder_model.onnx` | Encodes the input sentence |
+| `decoder_model.onnx` | Generates the translated tokens |
+| `source.spm` | Source language (German) tokenizer |
+| `target.spm` | Target language (Arabic) tokenizer |
+
+## Usage in Android
+
+Download the files and run inference with [ONNX Runtime for Android](https://onnxruntime.ai/).
+
+```
+Encoder input:  input_ids, attention_mask
+Encoder output: encoder_hidden_states
+
+Decoder input:  decoder_input_ids, encoder_hidden_states, encoder_attention_mask
+Decoder output: logits
+```
+
+Run the encoder once per sentence, then loop the decoder to generate tokens until the EOS token is produced.
+
+## Credits
+
+Original model by [Helsinki-NLP](https://huggingface.co/Helsinki-NLP), trained on [OPUS](https://opus.nlpl.eu/) data.  
+ONNX conversion by [YOUR_NAME].
+
+## License
+
+[CC-BY 4.0](https://creativecommons.org/licenses/by/4.0/) — same as the original model.