# how do I use dlpackrs with apache TVM for zero-copy?

**URL:** <https://discuss.tvm.apache.org/t/how-do-i-use-dlpackrs-with-apache-tvm-for-zero-copy/17658>\
**Category:** Questions\
**Created:** [August 28, 2024, 2:09pm UTC](https://discuss.tvm.apache.org/t/how-do-i-use-dlpackrs-with-apache-tvm-for-zero-copy/17658 "2024-08-28T14:09:03Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![tempmac](https://discuss.tvm.apache.org/letter_avatar_proxy/v4/letter/t/8e7dd6/32.png) [@tempmac](https://discuss.tvm.apache.org/u/tempmac)\
**Post date:** [August 28, 2024, 2:09pm UTC](https://discuss.tvm.apache.org/t/how-do-i-use-dlpackrs-with-apache-tvm-for-zero-copy/17658/1 "2024-08-28T14:09:03Z")

</div>

Hello!

is there a sample code for using dlpackrs ([https://github.com/ehsanmok/dlpackrs.git](https://github.com/ehsanmok/dlpackrs.git)) to convert tvm’s `NDArray` into `Array3<f32>` ?

context: I’m using YoloV5 converted to TVM. inference runs in microseconds, but I’m forced to pull the whole output from gpu to cpu (`output.to_vec()`) too early, which takes about 50~100 milliseconds.

thank you!