pytorch multiprocessing跳出进程之后无法获得梯度

最新推荐文章于 2024-10-05 20:00:00 发布

qq_28129141

最新推荐文章于 2024-10-05 20:00:00 发布

阅读量145

点赞数 9

文章标签： pytorch 人工智能 python

本文链接：https://blog.csdn.net/qq_28129141/article/details/140731286

版权

经过一段如下的多进程计算之后
for ith in range(nth):
th[ith] = Process( target=self.ComptTInteriorthread0, args=(geom,cfg,netith,min(netith+net,nnode),iz,solution,solz))

for ith in range(nth):
  th[ith].start()
for ith in range(nth):
  th[ith].join()

在每个进程内可以计算solution的导数，但是退出进程之后导数信息就不存在了，要如何让导数信息在跳出进程之后还能计算呢。

Traceback (most recent call last):
File “”, line 1, in
File “c:\Users\sphin\AppData\Local\Programs\Python\Python39\lib\site-packages\torch_tensor.py”, line 487, in backward
torch.autograd.backward(
File “c:\Users\sphin\AppData\Local\Programs\Python\Python39\lib\site-packages\torch\autograd_init_.py”, line 200, in backward
Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
RuntimeError: element 0 of tensors does not require grad and does not have a grad_fn