FunASR在310P上找不到芯片

一、问题现象（附报错日志上下文）：
需要用到ModelScope-FunASR来启动服务接入到dify，不知道咱们华为的平台用什么框架来调用。以下是模型列表：iic/speech_paraformer-large-vad-punc_asr_nat-zh-cn-16k-common-vocab8404-pytorch
iic/speech_fsmn_vad_zh-cn-16k-common-pytorch
iic/punc_ct-transformer_zh-cn-common-vocab272727-pytorch
iic/speech_campplus_sv_zh-cn_16k-common

二、软件版本:
torch版本2.1.0
python版本 3.9.5
驱动Version: 24.1.rc1
cann版本：8.0RC3
芯片 310P

三、测试步骤：
1.使用funasr提供的镜像：

sudo docker pull \
  registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.6
mkdir -p ./funasr-runtime-resources/models
sudo docker run -p 10095:10095 -it --privileged=true \
  -v $PWD/funasr-runtime-resources/models:/workspace/models \
  registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.6

2.创建容器：

docker run -it --name ljy_funasr_mars -p 10095:10095 \
--user root \
--net=host \
--privileged \
--device=/dev/davinci0 \
--device=/dev/davinci1 \
--device=/dev/davinci2 \
--device=/dev/davinci3 \
--device=/dev/davinci4 \
--device=/dev/davinci_manager \
--device=/dev/devmm_svm \
--device=/dev/hisi_hdc \
-v /usr/local/Ascend/driver:/usr/local/Ascend/driver \
-v /usr/local/Ascend/add-ons/:/usr/local/Ascend/add-ons\
-v /usr/local/sbin/npu-smi:/usr/local/sbin/npu-smi \
-v /data:/data \
-v /home:/home \
-v /var/log/npu:/usr/slog  \
registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.6 \
/bin/bash

3. 进入容器，安装cann和torch_npu:

3.1 安装cann：

Ascend-cann-toolkit_8.0.RC3_linux-aarch64.run --install
Ascend-cann-kernels-910_8.0.RC3_linux-aarch64.run --install
Ascend-cann-nnrt_8.0.RC3_linux-aarch64.run --install
下载链接：

https://www.hiascend.com/developer/download/community/result?module=ie+pt+cann&ie=1.0.RC3&pt=6.0.RC3.beta1&cann=8.0.RC3.beta1

3.2 安装torch_npu:
# 下载PyTorch安装包
wget https://download.pytorch.org/whl/cpu/torch-2.2.0-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
# 下载torch_npu插件包
wget https://gitee.com/ascend/pytorch/releases/download/v6.0.rc2-pytorch2.2.0/torch_npu-2.2.0.post2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
# 安装命令
pip3 install torch-2.2.0-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl
pip3 install torch_npu-2.2.0.post2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl

三、推理demo测试

1.下载模型权重：

parafomer热词版：https://www.modelscope.cn/models/iic/speech_paraformer-large-contextual_asr_nat-zh-cn-16k-common-vocab8404/files

vad语音端点检测：https://www.modelscope.cn/models/iic/speech_fsmn_vad_zh-cn-16k-common-pytorch/files

punc标点模型：https://www.modelscope.cn/models/iic/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727/files

paraformer在线版本：https://www.modelscope.cn/models/iic/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online/files

2.语音识别demo，cpu推理

from funasr import AutoModel
# paraformer-zh is a multi-functional asr model
# use vad, punc, spk or not as you need
model = AutoModel(model="/home/ljy/funasr/model_zoo/hot_para",  
                  vad_model="/home/ljy/funasr/model_zoo/vad", 
                  punc_model="/home/ljy/funasr/model_zoo/punc",
                  # spk_model="cam++"
                  )
res = model.generate(input="/home/ljy/funasr/tests/BAC009S0002W0122.wav",
            batch_size_s=300,
            hotword='魔搭')
print(res)

.demo是推理一条wav语音文件，输出检测得到的文字：
![输入图片说明](https://foruda.gitee.com/images/1749691376039994878/3468acdd_14180172.png "屏幕截图")

四、日志信息:

(base) root@ubuntu:~# docker run -p 10095:10095 -it --privileged=true \
>   -v $PWD/funasr-runtime-resources/models:/workspace/models \
>   registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.6
root@15ea4c56ced7:/workspace# npu-smi info
bash: npu-smi: command not found

Ascend/pytorch
暂停

内容风险标识

评论 (3)

Ascend/pytorch暂停 .gitee-modal { width: 500px !important; }

内容风险标识