# Running Multiple Python Interpreters via Custom Dynamic Loading

**URL:** <https://dev-discuss.pytorch.org/t/running-multiple-python-interpreters-via-custom-dynamic-loading/241>\
**Category:** Uncategorized\
**Created:** [May 25, 2021, 8:44pm UTC](https://dev-discuss.pytorch.org/t/running-multiple-python-interpreters-via-custom-dynamic-loading/241 "2021-05-25T20:44:52Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![zdevito](https://yyz2.discourse-cdn.com/flex036/user_avatar/dev-discuss.pytorch.org/zdevito/32/17_2.png) [@zdevito](https://dev-discuss.pytorch.org/u/zdevito)\
**Post date:** [May 25, 2021, 8:44pm UTC](https://dev-discuss.pytorch.org/t/running-multiple-python-interpreters-via-custom-dynamic-loading/241/1 "2021-05-25T20:44:52Z")

</div>

As part of torch::deploy, we’ve been prototyping a way of loading multiple Python interpreters into a single process using a custom shared library loader as a way to get around the global interpreter lock. With this approach many C extension libraries such as Numpy work out of the box.

> **[zdevito/custom\_loader](https://github.com/zdevito/custom_loader)**
>
> Contribute to zdevito/custom\_loader development by creating an account on GitHub.