| Confucius4-R2T2, Core ML / Apple Neural Engine conversion |
| ========================================================== |
|
|
| This is a Derivative Work of the NetEase Youdao Confucius4-R2T2 model. |
|
|
| Original model |
| NetEase Youdao Confucius4-R2T2 |
| Copyright (c) NetEase Youdao. All rights reserved. |
| https://huggingface.co/netease-youdao/Confucius4-R2T2 |
| https://github.com/netease-youdao/Confucius4-R2T2 |
| Released under the NetEase Youdao Model Use License Agreement (see MODEL_LICENSE). |
|
|
| Base model |
| Qwen3-ASR-1.7B, Copyright (c) Alibaba Cloud. |
| https://huggingface.co/Qwen/Qwen3-ASR-1.7B |
| Released under the Apache License, Version 2.0 (see LICENSE-Qwen3-ASR.txt). |
| The tokenizer files (tokenizer.json, tokenizer_config.json) are those of the original model. |
|
|
| This conversion |
| The weights in this repository are the original Confucius4-R2T2 weights converted for Core ML on |
| Apple silicon: the audio encoder in fp16, the text decoder palettised to 8 bits (LUT8) in two |
| stateful chunks, and the language-model head palettised to 8 bits. No re-training or fine-tuning |
| was performed. The conversion pipeline and the runtime that drives these files are published with |
| the VoiceInk fork that uses them (GPL-3.0, like VoiceInk). |
|
|
| Any modifications made to the original model in this Derivative Work are not endorsed, warranted, |
| or guaranteed by the original right-holder of the original model, and the original right-holder |
| disclaims all liability related to this Derivative Work. |
|
|
| Use of these files constitutes acceptance of the NetEase Youdao Model Use License Agreement |
| (MODEL_LICENSE). Anyone redistributing these files, or any further derivative, must keep this |
| NOTICE and the MODEL_LICENSE with every copy. |
|
|
