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.

Sign up for GitHub

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

adding weird issue #16

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

adding weird issue #16

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions docs/other.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,3 +7,7 @@ Stuff I or others have seen (anecdotal evidence):
- The A12 and later processors have a "smart compute system" that determines whether tasks should be run on the CPU, GPU, or ANE. So even if the model can run on the ANE without problems, Core ML may still decide to run it elsewhere, depending on what the system is doing. From [Twitter](https://twitter.com/eugenebokhan/status/1251423554861752320?s=20): "What we’ve also found out is that if NPU supports all the layers of the model and the GPU is resting, with a high probability CoreML will run it on GPU. But if GPU is heavy loaded with work - you can expect the model to be run on NPU."

- Even if all the layers in your model are compatible with the ANE, Core ML may still decide to run your model on the GPU. If you believe this is happening, split your model into two parts and see what happens to the first part. Does that run on the ANE now? I've seen it happen that, once the model becomes larger than a certain size, Core ML may decide that the GPU is more appropriate, even if the model could completely run on the ANE.

## Macbook Air vs Macbook Pro M1Max

Performance on densenet 121 is faster on Macbook Air than M1Max Macbook Pro. The difference is 4-5% with Monterey 12.3.1 and 13.3.1. Earlier version of Monterey and XCode had errors using ANE. The error was "H11ANEDevice::H11ANEDeviceOpen IOServiceOpen failed result= 0xe00002e2"