corey-nm
corey-nm
I'm going to be breaking the parts of this PR that can easily be merged into multiple smaller PRs.
Closing this for now - will reopen when work resumes
If this looks good I'll do pretty much the same thing in deepsparse
great questions! > what's the expected behavior on torch
Closing this and will be re-doing as: 1. Download models from sparsezoo 2. analyzer those. This will decouple it from torch versions and also test more edge cases
Can you add some unit tests to make sure this doesn't regress? Also is this a problem for other pipelines?
@glenn-jocher this use case came up with some quantization workflows that override the BaseTrainer class. So our custom subclasses were doing the fusing. I think the is_fused function should work...
> LGTM, just curious, what was the motivation for the change? This method of saving was decided for the standardization of integrations. It also makes it more clear when a...