# Chat with PDF answer model notice

pdfeditoronlinefree uses the unmodified quantized ONNX weights from `Xenova/flan-t5-small` at immutable revision `311454e83bc784267fd7eef5940ee854144abbec`.

The repository converts `google/flan-t5-small` revision `0fc9ddf78a1e988dac52e2dac162b0ede4fd74ab` for Transformers.js. The model is licensed under Apache-2.0.

- Encoder: `onnx/encoder_model_quantized.onnx`
- Encoder size: `35,759,127` bytes
- Encoder SHA-256: `968ce399b8dcdfd4c36833db51e9f78f327f7272ee9e2777fc2b3c8693208e90`
- Decoder: `onnx/decoder_model_merged_quantized.onnx`
- Decoder size: `59,339,331` bytes
- Decoder SHA-256: `73e2e942503221d7844715a8d824f68d8a2e878483f1c923849c177f0c441df1`
- Upstream ONNX model: https://huggingface.co/Xenova/flan-t5-small
- Base model: https://huggingface.co/google/flan-t5-small

The Apache-2.0 license text is retained at `/licenses/audio-to-pdf/xenova-transformers-2.17.2/LICENSE`.