# Create custom or retrain existing neural network for PYNQ? **URL:** \ **Category:** Support\ **Created:** [November 21, 2019, 11:15pm UTC](https://discuss.pynq.io/t/create-custom-or-retrain-existing-neural-network-for-pynq/686 "2019-11-21T23:15:01Z")\ **Posts on this page:** 2\ **Page:** 1
**Author:** ![nerdsley](https://avatars.discourse-cdn.com/v4/letter/n/779978/32.png) [@nerdsley](https://discuss.pynq.io/u/nerdsley)\ **Post date:** [November 21, 2019, 11:15pm UTC](https://discuss.pynq.io/t/create-custom-or-retrain-existing-neural-network-for-pynq/686/1 "2019-11-21T23:15:01Z")
I’m attempting to design a neural network using my own set of input data. I don’t mind using an existing network, just having the ability to train a new one. I’ve been exploring the QNN overlays for the PYNQ but it’s not clear to me where I should start looking to generate my own QNN weights. I’ve been using this as reference: > **[GitHub - Xilinx/QNN-MO-PYNQ](https://github.com/Xilinx/QNN-MO-PYNQ)** > > Contribute to Xilinx/QNN-MO-PYNQ development by creating an account on GitHub. Is there a way to create a new model or retrain an existing one? I’ve looked into generating a model through tensorflow but this may not deploy correctly. Thanks! ---
**Author:** ![gnatale](https://sea1.discourse-cdn.com/flex019/user_avatar/discuss.pynq.io/gnatale/32/11_2.png) [@gnatale](https://discuss.pynq.io/u/gnatale)\ **Post date:** [November 25, 2019, 11:58am UTC](https://discuss.pynq.io/t/create-custom-or-retrain-existing-neural-network-for-pynq/686/2 "2019-11-25T11:58:14Z")
The repo you have referenced is based upon the [FINN](https://xilinx.github.io/finn/) project, which is a really interesting project from a team of the Xilinx Research Labs. Since the kind of help you are asking is more orthogonal to PYNQ, and related to how to use FINN, I would suggest you to redirect your question to the team beyond it. You may want to check out their resources (from the website I linked) and as last resort open an issue on their github! Hope this helps.