# Saving ray model to tf/pytorch

**URL:** https://discuss.ray.io/t/saving-ray-model-to-tf-pytorch/11758
**Category:** Checkpointing, Restoring
**Created:** [August 11, 2023, 12:38pm UTC](https://discuss.ray.io/t/saving-ray-model-to-tf-pytorch/11758 "2023-08-11T12:38:18Z")
**Posts on this page:** 1
**Page:** 1

<div class="post-metadata">

### Author: ![Evangelista01](https://sea2.discourse-cdn.com/flex020/user_avatar/discuss.ray.io/evangelista01/32/4911_2.png) [@Evangelista01](https://discuss.ray.io/u/Evangelista01)
#### Post date: [August 11, 2023, 12:38pm UTC](https://discuss.ray.io/t/saving-ray-model-to-tf-pytorch/11758/1 "2023-08-11T12:38:18Z")

</div>

I would like to save ray model into something that could be loaded by pytorch/tensorflow for inference. Is it possible to avoid using ray serve and predict directly from the checkpoint files? Can you provide some code snippet on how to do it?