Views
No views yet
| 文件 | 说明 |
|---|---|
axmodel/encoder.axmodel / axmodel/decoder_loop.axmodel | Pulsar2 7.0-lite 重转换（encoder U16 + decoder U8 密集校准） |
fsmn_vad/fsmn_vad_10s_fp32.axmodel + fsmn_vad/am.mvn | FSMN-VAD（FunASR）NPU 语音门控 |
1conda create -n fireredasr python=3.12
2conda activate fireredasr
3pip install -r requirements.txt
4pip install axengine-0.1.3-py3-none-any.whl # pyaxengine
5python openai/openai_server.py \
6 --encoder axmodel/encoder.axmodel \
7 --decoder axmodel/decoder_loop.axmodel \
8 --fsmn-vad fsmn_vad/fsmn_vad_10s_fp32.axmodel \
9 --fsmn-cmvn fsmn_vad/am.mvn \
10 --cmvn axmodel/cmvn.ark --dict axmodel/dict.txt \
11 --pe axmodel/pe.npy --max-dur 10 --max-steps 128cpp/bin/firered_openai_server、cpp/bin/firered_openai_client
（用法见 cpp/bin/README.md）openai/openai_server.py、openai/openai_client.py（stdlib，零额外依赖）openai/firered_asr.py + openai/fsmn_vad_post.py
（numpy + pyaxengine + kaldi-native-fbank，无 torch/onnxruntime）POST /v1/audio/transcriptions（multipart file）、GET /v1/models