# PyTorch 1.10 is out!

**URL:** <https://community.wandb.ai/t/pytorch-1-10-is-out/1064>\
**Category:** Show the Community!\
**Created:** [October 21, 2021, 4:50pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064 "2021-10-21T16:50:59Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![bhutanisanyam1](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/bhutanisanyam1/32/18_2.png) [@bhutanisanyam1](https://community.wandb.ai/u/bhutanisanyam1)\
**Post date:** [October 21, 2021, 4:50pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064/1 "2021-10-21T16:50:59Z")

</div>

What are your favorite highlights from the new release? 😃

[Link to GH release](https://github.com/pytorch/pytorch/releases/tag/v1.10.0)

---

<div class="post-metadata">

**Author:** ![charlesfrye](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/charlesfrye/32/52_2.png) [@charlesfrye](https://community.wandb.ai/u/charlesfrye)\
**Post date:** [October 21, 2021, 5:37pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064/2 "2021-10-21T17:37:00Z")

</div>

> CUDA Graphs APIs are integrated to reduce CPU overheads for CUDA workloads.

This is awesome, even if it’ll take a while to bear fruit and will fly below most folks radar.

With [CUDA Graphs](https://pytorch.org/docs/master/generated/torch.cuda.CUDAGraph.html#torch.cuda.CUDAGraph), you can define some components of your model to effectively use TF1-style static execution, even while other components are dynamic.

I’ve been [profiling PyTorch code a lot more lately](http://wandb.me/trace-report), and you’d be surprised how often your code is CPU-bound, not GPU-bound, especially if you’re using smaller models.

My expectation is that eventually libraries will start to incorporate CUDA Graphs into their model implementations, so that end-users can get speedups without having to think about the CUDA layer at all.

IIRC, CUDA Graphs are capable of speeding up GPU executions by managing memory more effectively – think of it like making a “fused” layer, but not limited to only layers for which there are specific, handmade fused kernels.

---

<div class="post-metadata">

**Author:** ![cayush](https://avatars.discourse-cdn.com/v4/letter/c/c6cbf5/32.png) [@cayush](https://community.wandb.ai/u/cayush)\
**Post date:** [October 21, 2021, 6:04pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064/3 "2021-10-21T18:04:27Z")

</div>

I like the focus on mobile! Things like profilers for lite models, CoreML support, NNAPI, and others. I’d love to see PyTorch win in the lite/mobile/on-device segment

---

<div class="post-metadata">

**Author:** ![bhutanisanyam1](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/bhutanisanyam1/32/18_2.png) [@bhutanisanyam1](https://community.wandb.ai/u/bhutanisanyam1)\
**Post date:** [October 21, 2021, 6:22pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064/4 "2021-10-21T18:22:45Z")

</div>

> [@charlesfrye](#):
>
> This is awesome, even if it’ll take a while to bear fruit and will fly below most folks radar.

I definitely missed this one 😅

---

<div class="post-metadata">

**Author:** ![\_scott](https://sea2.discourse-cdn.com/flex020/user_avatar/community.wandb.ai/_scott/32/95_2.png) [@\_scott](https://community.wandb.ai/u/_scott)\
**Post date:** [October 22, 2021, 1:57pm UTC](https://community.wandb.ai/t/pytorch-1-10-is-out/1064/6 "2021-10-22T13:57:42Z")

</div>

> [@cayush](#):
>
> I love the focus on mobile. Things like profilers for lite models, CoreML support, NNAPI, and others.

+1 with this! [This tutorial on NNAPI](https://pytorch.org/tutorials/prototype/nnapi_mobilenetv2.html) makes it seem nice and simple to deploy models on Android devices.