Contribution for a test_code using lsq_lma on torch neural net modules

[xuesongwang]

Hi,
Thanks for providing this amazing project :)
While I was testing out the method to optimise neural network parameters written in torch, I realised the way to compute the Jacobian function was not straight forward. I managed to write a test demo that works.
I was wondering if you are open for contributions to put that in the test folder.
Cheers,

[xuesongwang]

Anyways, I will put my solution here :) https://github.com/xuesongwang/pytorch_Levenberg_Marquardt_optimizer