# Support for variable sharding / distributed inference?

**URL:** <https://discourse.edwardlib.org/t/support-for-variable-sharding-distributed-inference/131>\
**Category:** General\
**Created:** [May 17, 2017, 7:10am UTC](https://discourse.edwardlib.org/t/support-for-variable-sharding-distributed-inference/131 "2017-05-17T07:10:46Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![oscartackstrom](https://avatars.discourse-cdn.com/v4/letter/o/bb73d2/32.png) [@oscartackstrom](https://discourse.edwardlib.org/u/oscartackstrom)\
**Post date:** [May 17, 2017, 7:10am UTC](https://discourse.edwardlib.org/t/support-for-variable-sharding-distributed-inference/131/1 "2017-05-17T07:10:47Z")

</div>

Does Edward support sharding of the random variables and distributed inference? For example in LDA, one can imagine the word-topic matrix might not fit on a single machine if the number of topics/vocabulary is large.

---

<div class="post-metadata">

**Author:** ![dustin](https://yyz1.discourse-cdn.com/flex035/user_avatar/discourse.edwardlib.org/dustin/32/134_2.png) [@dustin](https://discourse.edwardlib.org/u/dustin)\
**Post date:** [May 23, 2017, 10:07pm UTC](https://discourse.edwardlib.org/t/support-for-variable-sharding-distributed-inference/131/2 "2017-05-23T22:07:41Z")

</div>

Edward is at the same abstraction level as TensorFlow. So I recommend looking into [Distributed TensorFlow](https://www.tensorflow.org/deploy/distributed), which is perfectly compatible.