# Unstable training in pytorch

**URL:** <https://discuss.pytorch.org/t/unstable-training-in-pytorch/19909>\
**Category:** Uncategorized\
**Created:** [June 19, 2018, 5:26am UTC](https://discuss.pytorch.org/t/unstable-training-in-pytorch/19909 "2018-06-19T05:26:05Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![alwynmathew](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/alwynmathew/32/8684_2.png) [@alwynmathew](https://discuss.pytorch.org/u/alwynmathew)\
**Post date:** [June 19, 2018, 5:26am UTC](https://discuss.pytorch.org/t/unstable-training-in-pytorch/19909/1 "2018-06-19T05:26:05Z")

</div>

I recently ported [monodepth](https://github.com/mrharicot/monodepth) which is written in tensorflow to pytorch([ported code](https://github.com/alwynmathew/monodepth-pytorch)). But the training seems unstable than the tensorflow implementation([github issue](https://github.com/alwynmathew/monodepth-pytorch/issues/1)).

For the first 10 - 20 epoch during training the result looks promising but its get very unstable after that. Could any one figure out the issue?