测试ONNX模型时有问题

[WjMessi1]

🐛 Bug

用测试onnx模型的样例代码测试本地某一文件夹下存储的相同类型音频时，有的音频可以，但是有的音频有报错。下面是我运行的样例代码： ` from funasr_onnx import Paraformer model_dir = "/home/data/FunASR_2024_3_26/shtel/models/finetune/kefu/823h/all_train_old_30000h_ep3_6000" model = Paraformer(model_dir, batch_size=1, quantize=True)

wav_path1 = ['/home/data/wavs/telecom/test-set-6h/split/530e0130-b746-4c8b-aa4f-93bcd124fd23_13901993778_1678023525-995675_agent.wav']

print("wav_path1: ",wav_path1) result1 = model(wav_path1) print("result1: ",result1)

wav_path2 = ['/home/data/wavs/telecom/test-set-6h/split/4808a180-2979-4120-b680-e350d95ae0c0_13818085208_1678066345-531642_agent.wav']

print("wav_path2: ",wav_path2) result2 = model(wav_path2) print("result2: ",result2) `

得到的输出：

Environment

• OS (e.g., Linux):
• FunASR Version (1.0.19):
• ModelScope Version (1.3.2):
• PyTorch Version (2.2.0):
• How you installed funasr (source):
• Python version: 3.8
• GPU (A100)

[WjMessi1]

更新了git下载链接，仓库里存放的是报错的模型文件（model_quant.onnx）以及音频文件（agent.wav）：

model_quant.onnx

agent.wav

[WjMessi1]

问题已经解决，环境的问题，谢谢