# How parametric Relu implemented internally **URL:** \ **Category:** Uncategorized\ **Created:** [July 23, 2019, 6:19pm UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421 "2019-07-23T18:19:17Z")\ **Posts on this page:** 5\ **Page:** 1
**Author:** ![dhakshiin](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/dhakshiin/32/12956_2.png) [@dhakshiin](https://discuss.pytorch.org/u/dhakshiin)\ **Post date:** [July 23, 2019, 6:19pm UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421/1 "2019-07-23T18:19:17Z")
i am trying to multiply the input vector of size (4,5,32,32) here 4 - batch size 5 - channels 32 x 32 feature size As mentioned in the parametric channelwise relu i want to multiply a vector to each of the channels ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [July 23, 2019, 10:06pm UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421/2 "2019-07-23T22:06:51Z")
Please don’t tag certain people specifically. They will answer, if they have time. Also, this might discourage other users to post an answer. ---
**Author:** ![dhakshiin](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/dhakshiin/32/12956_2.png) [@dhakshiin](https://discuss.pytorch.org/u/dhakshiin)\ **Post date:** [July 29, 2019, 3:25am UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421/3 "2019-07-29T03:25:28Z")
Sure. Can u pls provide suggestion for my question ---
**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\ **Post date:** [August 11, 2019, 9:37pm UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421/4 "2019-08-11T21:37:56Z")
Could you explain the parametric channelwise ReLU? How would you like to multiply the input of `[4, 5, 32, 32]` with a tensor, i.e. what shape would the internal tensor have? ---
**Author:** ![dhakshiin](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/dhakshiin/32/12956_2.png) [@dhakshiin](https://discuss.pytorch.org/u/dhakshiin)\ **Post date:** [August 18, 2019, 11:04am UTC](https://discuss.pytorch.org/t/how-parametric-relu-implemented-internally/51421/5 "2019-08-18T11:04:42Z")
The internal tensor will be of 1x channel number. For 5 channel input we would have 1x5 as the internal tensor. For [4 , 5, 32, 32] We need to multiply 1 value for each of the [32,32] in each batch. So 5 different values are used for each of the 5 channels in 32 x 32