模型部署在GPU，推理却还是用的CPU。求问

[Tian14267]

大神们， 我的模型是部署在GPU上。然后一共5条语音需要转写。结果前两条我看是用的GPU推理，后面几条语音，为啥又用CPU推理了？

我的代码；

import os
os.environ['CUDA_VISIBLE_DEVICES'] = '4'
from funasr import AutoModel

class ASRRecognize(object):
    def __init__(self):
        self.model = AutoModel(model="/data/fffan/0_experiment/15_ASR/1_code/0_models/0_funasr_models/speech_seaco_paraformer_large_asr_nat-zh-cn-16k-common-vocab8404-pytorch",model_revision="v2.0.4",
                          vad_model="/data/fffan/0_experiment/15_ASR/1_code/0_models/0_funasr_models/speech_fsmn_vad_zh-cn-16k-common-pytorch",vad_model_revision="v2.0.4",
                          punc_model="/data/fffan/0_experiment/15_ASR/1_code/0_models/0_funasr_models/punc_ct-transformer_zh-cn-common-vocab272727-pytorch",punc_model_revision="v2.0.4",
                          spk_model="/data/fffan/0_experiment/15_ASR/1_code/0_models/0_funasr_models/speech_campplus_sv_zh-cn_16k-common", spk_model_revision="v2.0.2",
                          device="cuda:0"
                          )

    def _recognize(self, voice_path,hotword):
        try:
            res = self.model.generate(input=voice_path,
                        batch_size_s=300,
                        hotword=hotword)

            res_all = {"text": res}
        except:
            res_all = {"text":None}
        return res_all

if __name__ == "__main__":
    ASR = ASRRecognize()
    for one_wav in wav_path_list:
        out = ASR._recognize(one_wav, "")

求助大神们帮忙看看

[liuqijie6]

我也遇到了这个问题，求解

[yeyunu]

结局了吗