How to use channel pruning on pre trained resnet 18?

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Dear Moderators,

I want your kind attention regarding the pruning methods. I want channel pruning for the ResNet18 pre-tarined model. How can I prune the layers which are not working or helpful during training so I can amend the model to get maximum efficiency/accuracy.
I am attaching a picture below regarding a pruning illustration of the algorithm/model.

thanks

dfggd
dfggd932×259 144 KB

3

Hi,

Please check our TAO toolkit below.
It can help to prune the non-necessary weight/layer among the training.

Thanks.

4

Dear, Will this toolkit work on Jetson Nano 2gb? (In regards to specs, capabilities… etc)

5

Hi,

Please apply training on a desktop GPU.
Then you can copy the model to Nano and deploy it with Deepstream.

Thanks.

7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.