Yup, ideally that will be the use case. Another thing to note is that it looks like nvidia is going to come up with an UI for TAO too
I am just trying to set up this repository of AI use cases for my team so that members can quickly produce a model on their limited custom dataset
Hmm interesting, I guess once you are able to connect it with ClearML you can just clone / modify / enqueue and let users train models directly from the UI on any hardware, is that the plan ?
it looks like nvidia is going to come up with an UI for TAO too
Interesting, any reference we could look at ?
I tried running TAO, what the cli does is that it spins up docker containers to do the actual training, which would complicate the clearml integration
that's why I think the clearml TLT example will no longer work with the latest TAO
I am not sure the difference in terms of triton serving, but in terms of retraining, I see from the ClearML TLT example code it relies on the nvidia IVA python package. The latest TAO doesn't use python for fine tuning, rather it uses the CLI entirely
Can you please expand regarding this new TAO and what's the difference to how triton serves at the moment?
The latest TAO doesn't use python for fine tuning, rather it uses the CLI entirely
It's a good question, but I think the CLI actually just runs a python code (the CLI is their interface). Generally speaking I'm pretty sure it will not be complicated to convert the TLT integration to support TAO (Nvidia helps with that, and I think we had a similar proces with Nvidia Clara/MONAI)
BTW: how are you using Nvidia TAO ?