HI, do we have auto-tune logs that can be used directly to build our models?

1

Hi,

Does community provide logs with which we can compile our models ? I cannot do autotune on my target platform, so I cannot compile my model to run it efficiently on my target platform. Does the community have these logs ?

2

Hi @coincheung ,

I guess this might be what you are looking for? Not sure how you should use it though.

Best regards

3

Hi,

Thanks for replying so soon !!!

So does tvm only tune according to libs such as cuda, rather than specific hardwares such as v100 and 2080ti ?

I mean to deploy my model on v100 server, but I do not have v100 at my hand, so I cannot tune my model locally. That is why I want to use the logs that is pre-tuned by other people, with which I can compile my model locally and deploy it directly on target platform. Is this idea feasible ?