# Initialize a new network from a sub-network

**URL:** <https://discuss.pytorch.org/t/initialize-a-new-network-from-a-sub-network/500>\
**Category:** Uncategorized\
**Created:** [February 14, 2017, 10:32pm UTC](https://discuss.pytorch.org/t/initialize-a-new-network-from-a-sub-network/500 "2017-02-14T22:32:52Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![pronics2004](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/pronics2004/32/137_2.png) [@pronics2004](https://discuss.pytorch.org/u/pronics2004)\
**Post date:** [February 14, 2017, 10:32pm UTC](https://discuss.pytorch.org/t/initialize-a-new-network-from-a-sub-network/500/1 "2017-02-14T22:32:52Z")

</div>

What is the best way to initialize a new network from a sub-network of larger network (saved using torch.save(net.state\_dict())?

---

<div class="post-metadata">

**Author:** ![fmassa](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/fmassa/32/58_2.png) [@fmassa](https://discuss.pytorch.org/u/fmassa)\
**Post date:** [February 14, 2017, 10:38pm UTC](https://discuss.pytorch.org/t/initialize-a-new-network-from-a-sub-network/500/2 "2017-02-14T22:38:14Z")

</div>

You could use some things from [How to extract features of an image from a trained model](https://discuss.pytorch.org/t/how-to-extract-features-of-an-image-from-a-trained-model/119/2)