# API-level Stateful RNNs?

**URL:** <https://discuss.pytorch.org/t/api-level-stateful-rnns/73815>\
**Category:** Uncategorized\
**Created:** [March 19, 2020, 11:30pm UTC](https://discuss.pytorch.org/t/api-level-stateful-rnns/73815 "2020-03-19T23:30:00Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![OverLordGoldDragon](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/overlordgolddragon/32/21887_2.png) [@OverLordGoldDragon](https://discuss.pytorch.org/u/OverLordGoldDragon)\
**Post date:** [March 19, 2020, 11:30pm UTC](https://discuss.pytorch.org/t/api-level-stateful-rnns/73815/1 "2020-03-19T23:30:00Z")

</div>

Does PyTorch support a Keras-like `LSTM(stateful=True)`, or it needs to be custom-coded? I’m a TensorFlow user, considering PyTorch, and so far PyTorch’s neural network API appears quite lacking relative to Keras (e.g. [this entire list](https://github.com/pytorch/pytorch/issues/9572), most of which Keras supports, also weight constraints).