# Failed call to cuInit: CUDA\_ERROR\_NOT\_INITIALIZED: initialization error

**URL:** <https://discuss.pytorch.org/t/failed-call-to-cuinit-cuda-error-not-initialized-initialization-error/167096>\
**Category:** vision\
**Created:** [November 28, 2022, 6:51pm UTC](https://discuss.pytorch.org/t/failed-call-to-cuinit-cuda-error-not-initialized-initialization-error/167096 "2022-11-28T18:51:30Z")\
**Posts on this page:** 3\
**Page:** 1

<div class="post-metadata">

**Author:** ![HamidGadirov](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/hamidgadirov/32/70216_2.png) [@HamidGadirov](https://discuss.pytorch.org/u/HamidGadirov)\
**Post date:** [November 28, 2022, 6:51pm UTC](https://discuss.pytorch.org/t/failed-call-to-cuinit-cuda-error-not-initialized-initialization-error/167096/1 "2022-11-28T18:51:30Z")

</div>

Once I am getting next batch from DataLoader, there is an error message:

`E tensorflow/compiler/xla/stream_executor/cuda/cuda_driver.cc:267] failed call to cuInit: CUDA_ERROR_NOT_INITIALIZED: initialization error`

The program finally crashes with `RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR` once the code reaches neural net. I have recently updated tensorflow to 2.11.0, PyTorch is 1.13.0+cu117, and I have NVIDIA TITAN V GPU with `nvidia-smi`:

 ![Screenshot 2022-11-28 at 19.49.29](https://discuss.pytorch.org/uploads/default/original/3X/8/3/8327add87639d4d8a35edf764715534cd5f6cf19.png)  
I am on conda environment, I have `cudnn 7.6.5.32` and `nvidia-cudnn-cu11 8.5.0.96` installed. Any help would be appreciated!

---

<div class="post-metadata">

**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\
**Post date:** [November 29, 2022, 12:38am UTC](https://discuss.pytorch.org/t/failed-call-to-cuinit-cuda-error-not-initialized-initialization-error/167096/2 "2022-11-29T00:38:30Z")

</div>

It seems you are running into this error while using TensorFlow, so their discussion board might be a better place to ask.

> [@HamidGadirov](#):
>
> E tensorflow/compiler/xla/stream\_executor/cuda/cuda\_driver.cc:267] failed call to cuInit: CUDA\_ERROR\_NOT\_INITIALIZED: initialization error

---

<div class="post-metadata">

**Author:** ![HamidGadirov](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/hamidgadirov/32/70216_2.png) [@HamidGadirov](https://discuss.pytorch.org/u/HamidGadirov)\
**Post date:** [November 29, 2022, 9:59am UTC](https://discuss.pytorch.org/t/failed-call-to-cuinit-cuda-error-not-initialized-initialization-error/167096/3 "2022-11-29T09:59:32Z")

</div>

Indeed, the problem is related to TensorFlow, if I remove it training starts without issues. I also asked on the TensorFlow forum.