File size: 3,521 Bytes
edcd0cc | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 | ---
license: other
license_name: netease-youdao-model-use-license-agreement
license_link: https://raw.githubusercontent.com/netease-youdao/Confucius4-R2T2/refs/heads/master/MODEL_LICENSE
base_model:
- netease-youdao/Confucius4-R2T2
pipeline_tag: automatic-speech-recognition
language:
- zh
- en
tags:
- FastFlowLM
- NPU
- AMD
- Ryzen-AI
- asr
- speech-recognition
- qwen3-asr
---
# Confucius4-R2T2 for AMD Ryzen™ AI NPUs (FastFlowLM)
[Confucius4-R2T2](https://huggingface.co/netease-youdao/Confucius4-R2T2) by NetEase Youdao ("Real Real-Time Transcription", fine-tuned from
[Qwen3-ASR-1.7B](https://huggingface.co/Qwen/Qwen3-ASR-1.7B)) converted for [FastFlowLM](https://github.com/FastFlowLM/FastFlowLM) (FLM) on the XDNA 2 NPU.
2B parameters, 30 languages (best at Chinese and English).
> 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.
This is a **converted copy** of the original weights, subject to the **NetEase Youdao Model Use License Agreement** (`MODEL_LICENSE`, included). Please read it: among other
terms it requires a separate license from NetEase Youdao for organizations with more than 100 million monthly active users or more than RMB 1 billion annual revenue
(including their affiliates), and it forbids use of the model or its outputs to improve other AI models except non-commercial ones.
## Run
Requires FastFlowLM 1.0.14 or newer.
```shell
flm run r2t2:1.7b
# then: /input "audio.wav" (wav, flac, mp3, ogg)
# /input "audio.wav" hot words, names, topic (text after the file is context for the model)
flm serve r2t2:1.7b
curl http://localhost:52625/v1/audio/transcriptions -F file=@audio.wav -F model=r2t2:1.7b
```
Whole-file transcription, or streaming with `-F stream=true` (append-only text while the audio is processed, `chunk_size_ms` 80-2000).
See the [model page](https://fastflowlm.com/docs/models/r2t2/) and the [benchmarks](https://fastflowlm.com/docs/benchmarks/r2t2_results/).
## What is in this package
| File | Content |
|---|---|
| `model.q4nx` | text decoder (Qwen3-1.7B shape): **Q8_0** projections and output head, BF16 embedding (the CPU prompt prefill reads it) |
| `npu_q41.q4nx` | **Q4_1** copies of q/k/o/gate/up and the output head for the NPU decode (v and down stay Q8_0 from `model.q4nx`) |
| `audio_weight.q4nx` | audio encoder (24 layers), BF16, in the NPU matmul tile layout |
| `config.json`, `tokenizer.json`, `tokenizer_config.json`, `chat_template.jinja` | from the original checkpoint |
| `MODEL_LICENSE` | NetEase Youdao Model Use License Agreement (original) |
The NPU decode reads mixed precision: 4-bit everywhere (Q4_0, Q4_1, Q4_K round-to-nearest) breaks this fine-tune (empty
transcripts or Mandarin answered in English), but Q4_1 for q/k/o/gate/up and the output head with v_proj and down_proj at Q8_0
keeps its output - and decodes ~24 % faster than Q8_0 throughout. On 120 English (LibriSpeech test-clean) and 60 Mandarin
(FLEURS) utterances, compared with the original BF16 model, FLM gives WER 2.81 % vs 2.89 % (English) and CER 6.37 % vs 8.96 %
(Mandarin; the gap is mostly how numbers are written, digits vs Chinese numerals, which the scoring does not normalise).
## Source
Converted from `netease-youdao/Confucius4-R2T2` (BF16, revision `185ce639`) with the FastFlowLM q4nx converter.
|