# Why DataParallel works poorly for CNNs compared to LSTMs?

**URL:** <https://discuss.pytorch.org/t/why-dataparallel-works-poorly-for-cnns-compared-to-lstms/25257>\
**Category:** Uncategorized\
**Created:** [September 16, 2018, 12:13am UTC](https://discuss.pytorch.org/t/why-dataparallel-works-poorly-for-cnns-compared-to-lstms/25257 "2018-09-16T00:13:20Z")\
**Posts on this page:** 1\
**Showing post:** 11

<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:** [September 16, 2018, 10:07pm UTC](https://discuss.pytorch.org/t/why-dataparallel-works-poorly-for-cnns-compared-to-lstms/25257/11 "2018-09-16T22:07:03Z")

</div>

Have a look at [this explanation](https://discuss.pytorch.org/t/debugging-dataparallel-no-speedup-and-uneven-memory-allocation/1100/14?u=ptrblck) of the general parallel operations.

---

_[View the full topic](https://discuss.pytorch.org/t/why-dataparallel-works-poorly-for-cnns-compared-to-lstms/25257)._