rt,autodl的3080ti12g显存,把batchsize降到4才能开始跑,然后跑一轮在验证的时候爆,其他的batch直接是跑不了一点![](http://static.tieba.baidu.com/tb/editor/images/client/image_emoticon16.png)
报错是这样:
CUDA out of memory. Tried to allocate 2.97 GiB (GPU 0; 11.76 GiB total capacity; 6.23 GiB already allocated; 2.45 GiB free; 7.37 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
![](http://tiebapic.baidu.com/forum/w%3D580/sign=a83641c1e1efce1bea2bc8c29f50f3e8/854c64df8db1cb13718d3e249b54564e93584b01.jpg?tbpicau=2024-09-23-05_66cdddb5a577ef5e6cad297d86ca8e1a)
![](http://tiebapic.baidu.com/forum/w%3D580/sign=f5c2aa11bbfaaf5184e381b7bc5594ed/7af773540923dd546d97a55f9709b3de9d824801.jpg?tbpicau=2024-09-23-05_4328798b3c2e73c992813f156adb833e)
![](http://tiebapic.baidu.com/forum/w%3D580/sign=9d36f91cdb16fdfad86cc6e6848e8cea/3c55f11349540923e78d6149d458d109b2de4901.jpg?tbpicau=2024-09-23-05_183328bb2257ee455be52ff0f8295b71)
![](http://static.tieba.baidu.com/tb/editor/images/client/image_emoticon16.png)
报错是这样:
CUDA out of memory. Tried to allocate 2.97 GiB (GPU 0; 11.76 GiB total capacity; 6.23 GiB already allocated; 2.45 GiB free; 7.37 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
![](http://tiebapic.baidu.com/forum/w%3D580/sign=a83641c1e1efce1bea2bc8c29f50f3e8/854c64df8db1cb13718d3e249b54564e93584b01.jpg?tbpicau=2024-09-23-05_66cdddb5a577ef5e6cad297d86ca8e1a)
![](http://tiebapic.baidu.com/forum/w%3D580/sign=f5c2aa11bbfaaf5184e381b7bc5594ed/7af773540923dd546d97a55f9709b3de9d824801.jpg?tbpicau=2024-09-23-05_4328798b3c2e73c992813f156adb833e)
![](http://tiebapic.baidu.com/forum/w%3D580/sign=9d36f91cdb16fdfad86cc6e6848e8cea/3c55f11349540923e78d6149d458d109b2de4901.jpg?tbpicau=2024-09-23-05_183328bb2257ee455be52ff0f8295b71)