# I cannot use the pytorch that was built successfully from source: (DLL) initialization routine failed. Error loading caffe2\_detectron\_ops\_gpu.dll

**URL:** <https://discuss.pytorch.org/t/i-cannot-use-the-pytorch-that-was-built-successfully-from-source-dll-initialization-routine-failed-error-loading-caffe2-detectron-ops-gpu-dll/93243>\
**Category:** windows\
**Created:** [August 18, 2020, 3:03pm UTC](https://discuss.pytorch.org/t/i-cannot-use-the-pytorch-that-was-built-successfully-from-source-dll-initialization-routine-failed-error-loading-caffe2-detectron-ops-gpu-dll/93243 "2020-08-18T15:03:26Z")\
**Posts on this page:** 1\
**Showing post:** 9

<div class="post-metadata">

**Author:** ![InTheYear2525](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/i/db5fbb/32.png) [@InTheYear2525](https://discuss.pytorch.org/u/InTheYear2525)\
**Post date:** [August 26, 2020, 9:26am UTC](https://discuss.pytorch.org/t/i-cannot-use-the-pytorch-that-was-built-successfully-from-source-dll-initialization-routine-failed-error-loading-caffe2-detectron-ops-gpu-dll/93243/9 "2020-08-26T09:26:04Z")

</div>

I have stopped any tests on this after putting some days into this. I am waiting for a helping reply, good that you join me in waiting. Generally, I guess that the error is linked with the same problem that causes ninja not to work. I do not know why I cannot use ninja. Setting off ninja is just a workaround, yet there should be a better way I think, because ninja will make the install much faster. I do not want to let it run for 7 hours again and again only to have the same errors thrown.

If you want to try something that seems promising, have a look at [PyTorch build from source on Windows](https://discuss.pytorch.org/t/pytorch-build-from-source-on-windows/40288/16) and further down, especially

`set CUDAHOSTCXX=`

---

_[View the full topic](https://discuss.pytorch.org/t/i-cannot-use-the-pytorch-that-was-built-successfully-from-source-dll-initialization-routine-failed-error-loading-caffe2-detectron-ops-gpu-dll/93243)._