Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

When will aimet support pytorch2.x #2869

Open
buxianggaimingzi opened this issue Apr 10, 2024 · 7 comments
Open

When will aimet support pytorch2.x #2869

buxianggaimingzi opened this issue Apr 10, 2024 · 7 comments

Comments

@buxianggaimingzi
Copy link

Is there any plan or time for support pytorch2.x? Thank you very much.

@escorciav
Copy link

Isn't the pytorch.compile the go to way in 2.x?
If so, it might imply a big departure wrt to torch.fx.

My bet: AIMET might be deprecated or get its scope reduced in 3-5 years

@quic-mangal
Copy link
Contributor

@buxianggaimingzi, we are currently working on supporting PT2.x. Will keep you posted regarding the update.

@escorciav
Copy link

Good to know.
Cheering up for the team 💪

Looking fwd to hearing more news.
Especially related to Executorch.

@escorciav
Copy link

escorciav commented May 13, 2024

@buxianggaimingzi out of curiosity, did you try building/compiling AIMET for PyTorch >= 2.0?
@quic-mangal kindly let us know if building/compilation is a dead end with PyTorch 2.0 or above

@quic-mangal
Copy link
Contributor

@escorciav, currently we are still working on the upgrade. Hopefully it should be available by EOM.

@escorciav
Copy link

@quic-mangal I understood that. However, I was wondering if I can still build AIMET from source with PyTorch >= 2.1.

A friend did it (fork). However, I'm unable to reproduce his recipe. I'll open another issue to not derail the convo :)

@escorciav
Copy link

escorciav commented May 21, 2024

ping AIMET:maintainers/contributors/users

My friend's fork is a bit outdated. I have to update the model prep again.

Venting. I wish Executorch finally levels up PyTorch on-device & at-the-edge. Thus, AIMET scope (for PyTorch users) gets reduced to (1) kernel delegates (plugins) & recipes, + (2) legacy tool. The latter is mostly for Qualcomm folks. You know as Tensorflow brought up Theano early retirement. I often feel as 👇
image

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants