RuntimeError: CUDA out of memory. Tried to allocate 5. If reserved memory is ＞＞ allocated memory_if reserved memory is >> allocated memory try sett

作者：AllinToyou | 2024-05-05 10:13:11

踩

if reserved memory is >> allocated memory try setting max_split_size_mb to a

报错信息如下：

RuntimeError: CUDA out of memory. Tried to allocate 50.00 MiB (GPU 0; 5.80 GiB total capacity; 4.39 GiB already allocated; 35.94 MiB free; 4.46 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

经过向ChatGPT询问，得到如下答案：


import torch
torch.cuda.set_per_process_memory_fraction(0.9, 0)  # 可根据需要调整0.9的值
torch.cuda.set_per_process_memory_growth(True)
 
 
torch.cuda.set_max_split_size(max_size_mb=1024)  # 根据需要调整max_size_mb的值

声明：本文内容由网友自发贡献，不代表【wpsshop博客】立场，版权归原作者所有，本站不承担相应法律责任。如您发现有侵权的内容，请联系我们。转载请注明出处：https://www.wpsshop.cn/w/AllinToyou/article/detail/538478