Paddle2Torch问题咨询

[superdonkey007]

@cangtianhuang
你好，咨询一下按照教程进行

环境配置时，paddle和pytorch所依赖的nvidia-cudnn-cu11 和nvidia-nccl-cu11的版本冲突怎么处理？

[cangtianhuang]

这两个框架会有部分文件冲突，请尝试先安装 paddlepaddle，再安装 pytorch，这样可以避免冲突

重新安装时，请尝试添加 --force-reinstall 参数

[superdonkey007]

使用aistudio的配置进行测试时，经常遇到GPU内存不足的问题，如：
[test error] paddle.linalg.norm(Tensor([11408507, 10, 20],"float32"), p=2.0, axis=-1, ) CUDA out of memory. Tried to allocate 8.50 GiB. GPU 0 has a total capacity of 31.74 GiB of which 1.88 GiB is free. Process 1074311 has 29.86 GiB memory in use. Of the allocated memory 28.05 GiB is allocated by PyTorch, and 6.00 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)
这个只有重新租赁设备一种解决办法吗？

[cangtianhuang]

@superdonkey007 paddle.linalg.norm(Tensor([11408507, 10, 20],"float32"), p=2.0, axis=-1, ) 配置属于超大 tensor，这个我们暂不方便做精度对比。仅需测试所有 tester/api_config/api_config_merged_*.txt 内的常规配置即可～😉