# 'aten::native\_batch\_norm' with arguments from the 'SparseCUDA' backend

**URL:** <https://discuss.pytorch.org/t/aten-native-batch-norm-with-arguments-from-the-sparsecuda-backend/130020>\
**Category:** Uncategorized\
**Created:** [August 23, 2021, 6:51am UTC](https://discuss.pytorch.org/t/aten-native-batch-norm-with-arguments-from-the-sparsecuda-backend/130020 "2021-08-23T06:51:53Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![pytorching](https://discuss.pytorch.org/letter_avatar_proxy/v4/letter/p/85f322/32.png) [@pytorching](https://discuss.pytorch.org/u/pytorching)\
**Post date:** [August 23, 2021, 6:51am UTC](https://discuss.pytorch.org/t/aten-native-batch-norm-with-arguments-from-the-sparsecuda-backend/130020/1 "2021-08-23T06:51:53Z")

</div>

Does `torch.nn.batch_norm` have support for sparse tensor?  
In my test:

 ![image](https://discuss.pytorch.org/uploads/default/original/3X/9/1/914c04226d400f58949824d5a12db5be2e59e6f7.png)

Both ‘SparseCPU’ 和 ‘SparseCUDA’ backend doesn’t support.