Congi-libya/BayanSimplify-ONNX

Bayan on-device models (ONNX, int8)

The simplification models the Bayan Android app downloads. Each folder is one bundle: encoder.onnx and decoder.onnx (int8 dynamic quantization, merged decoder with KV cache), tokenizer.json, and bayan_model.json (prefix, decoding settings and the decoder's inputs and outputs). The app checks every file's SHA-256 before using it.

folder	model	size
model3/	model 3: AraT5v2 fine-tuned on the Bayan corpus v1; the app's "Large model"	471 MB
arabart/	model 2, AraBART; the app's "Fast model"	222 MB
arat5/	model 2, AraT5v2; the app's previous "Large model"	471 MB

In the app, each output passes through its text step: scripture and set poetry are never rewritten, and a sentence keeps its source wording when a rewrite drops a number, a Latin-script word, or changes negation or limits.

Licence: CC BY-NC 4.0 (non-commercial), in line with the training data's terms.