libgomp: Thread creation failed: Resource temporarily unavailable

[loxs123]

What is your question?

使用多次调用同一个模型报错，首次调用generate并不会报错，后2次或者3次就会报错

libgomp: Thread creation failed: Resource temporarily unavailable

libgomp: 
libgomp: 
libgomp: Thread creation failed: Resource temporarily unavailableThread creation failed: Resource temporarily unavailableThread creation failed: Resource temporarily unavailable

libgomp: 
libgomp: Thread creation failed: Resource temporarily unavailableThread creation failed: Resource temporarily unavailable
libgomp: 

Thread creation failed: Resource temporarily unavailable

libgomp: Segmentation fault (core dumped)

Code

def worker_process():
    model= AutoModel(model="paraformer-zh",  vad_model="fsmn-vad",  punc_model="ct-punc", spk_model="cam++")
    wav_files = glob.glob('data/*/*/*.wav')
    for wav_file in wav_files:
        print('processing ',wav_file)
        output_path =f'{wav_file[:-4]}.json'
        start_time = time.time()
        ans = model.generate(input=wav_file,batch_size_s=300,hotword='', )
        os.system(f'rm "{wav_file}"')
        end_time = time.time()
        with open(output_path, 'w') as f:
            json.dump(ans, f, ensure_ascii=False)


if __name__ == '__main__':
    worker_process()

What's your environment?

OS: CentOS Linux release 7.9.2009
FunASR Version: 1.0.27
ModelScope Version: 1.14.0
PyTorch Version: 2.1.2+cu118
How you installed funasr: pip
Python version: 3.8.19
GPU: v100
CUDA/cuDNN version: 11.4
cpu

Architecture:          x86_64
CPU op-mode(s):        32-bit, 64-bit
Byte Order:            Little Endian
CPU(s):                64
On-line CPU(s) list:   0-63
Thread(s) per core:    2
Core(s) per socket:    32
Socket(s):             1
NUMA node(s):          1
Vendor ID:             GenuineIntel
CPU family:            6
Model:                 85
Model name:            Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Stepping:              4
CPU MHz:               2394.374
BogoMIPS:              4788.74
Hypervisor vendor:     KVM
Virtualization type:   full
L1d cache:             32K
L1i cache:             32K
L2 cache:              1024K
L3 cache:              28160K
NUMA node0 CPU(s):     0-63

[loxs123]

目前解决方案

def worker_process():
    wav_files = glob.glob('data/*/*/*.wav')
    for wav_file in wav_files:
        print('processing ',wav_file)
        output_path =f'{wav_file[:-4]}.json'
        start_time = time.time()
        model= AutoModel(model="paraformer-zh",  vad_model="fsmn-vad",  punc_model="ct-punc", spk_model="cam++")
        ans = model.generate(input=wav_file,batch_size_s=300,hotword='', )
        os.system(f'rm "{wav_file}"')
        end_time = time.time()
        with open(output_path, 'w') as f:
            json.dump(ans, f, ensure_ascii=False)


if __name__ == '__main__':
    worker_process()

[hopofb118]

目前只能不断初始化么，有没有办法可以维持这个实例啊

[loxs123]

目前只能不断初始化么，有没有办法可以维持这个实例啊

后来没研究了 我也很期待修复这个bug