quanzhang2020

模型合并时报错，ValueError: weight is on the meta device, we need a `value` to put in on cpu.

1


语音分类example运行报错

1

按照https://github.com/PaddlePaddle/PaddleSpeech/tree/r1.4/examples/esc50中描述，进行模型1、训练/2、推理/3、模型导出/4、导出模型推理运行，前三步都能成功，运行第四步报错如下：  环境： ubuntu18 python3.8 absl-py 2.1.0 aiohttp 3.9.5 aiosignal 1.3.1 aistudio-sdk 0.2.4 annotated-types 0.7.0 antlr4-python3-runtime 4.9.3 anyio 4.4.0 astor 0.8.1 asttokens 2.4.1 async-timeout 4.0.3 attrs 23.2.0 audioread 3.0.1 av...

同样docker容器，同样的代码3090/4070显卡报错，V100/A100训练成功

1

### bug描述 Describe the Bug https://github.com/yeyupiaoling/AudioClassification-PaddlePaddle 这个工程，在V100显卡机器按照相关的依赖安装，训练成功后，打包成docker；commit docker image导出到另外一台4070显卡机器，同样的样本，同样的参数运行会报cuda错误。各种库依赖详见：https://github.com/yeyupiaoling/AudioClassification-PaddlePaddle ### 其他补充信息 Additional Supplementary Information _No response_

'CUFFT_INTERNAL_ERROR'. Driver or internal cuFFT library error

1

env： cuda11.7 paddle2.6 gpu：3090 running on V100 is ok, but runing on GeForce RTX 4070 is wrong like this: File "/usr/local/lib/python3.10/dist-packages/paddleaudio/compliance/kaldi.py", line 468, in fbank spectrum = paddle.fft.rfft(strided_input).abs() File "/usr/local/lib/python3.10/dist-packages/paddle/fft.py",...