Model runs unefficienty after pruning

I train a model and prune it.
I run it in my xavier but founds that it runs slow a lot .
so why this happens

Hi lc5104hit,

Pleases maximize the device performance with the following commands first:

$ sudo nvpmodel -m 0

$ sudo jetson_clocks

does not work.
Is there any rules that I should follow?

I really puzzle about this result


Have you tried the non-pruned model on Xavier first?
Is it also slow?

If yes, the performance issue should originate from the frameworks you used.
Please share which framework do you use for inference that we can give you a further suggestion.

If not, there is some issue in your pruning process.
Please share the tool or source you used for the pruning.


the non-pruned model in xavier is about 70ms,
actually the pruned model in xavier is about 150ms

I think may exist some bug in trtexec tool.
I write inference code myself and found the model speed is okey now.

Thanks for your kindly reply.

best wished


Could you share the non-pruned and pruned model with us?
We want to reproduce this internally.


I cannot sent a file larger than 4mb,


Could you update to a third-party drive and share the link with us?


Hi lc5104hit

Is this still an issue to support?
Could you update to a third-party drive and share the link with us?


I try baiduyun


Do you have the link of baiduyun?


链接: 密码: 6c7q