Matthew Waller comments

Results 37 comments of


                                            Matthew Waller

Make model convertible by CoreML

Oops, Yep. Here it is @lissyx and @kdavis-mozilla https://developers.googleblog.com/2017/12/announcing-core-ml-support.html

reconstruction error

Ha! I ran into this issue too. Couldn’t make reconstruction work in the sample config setup. Also tried removing the y values and ran into the size mismatch. Hmmm

Using speaker diarization on mobile devices

Ah! So you're saying that the inputs are basically mfccs? I actually wrote an mfcc calculator in Swift in the style of python_speech_features. So should I be able to take,...

Using speaker diarization on mobile devices

Interesting. So looking at the `speakerDiarization.py` code more closely, would I need to import two models into the iOS or Android app? Looks like I would load one model here:...

Possible to use spec to break large model into smaller model?

You know, I did compress the weights, but everything still balloons to 8 GB. So it's more the operation going on inside the model, building up a ton of tensors....

Possible to use spec to break large model into smaller model?

Thanks @DawerG It's an mlprogram.

Possible to use spec to break large model into smaller model?

Thanks y'all, looks like breaking up the model is the way to go. Appreciate the insight!

Possible to use spec to break large model into smaller model?

@Lukas1h hey sure thing! I'm still working on breaking it down to the point that I'm able to use the ANE, but I intend to have a massive blog post...

Possible to use spec to break large model into smaller model?

@Lukas1h it does! It's not the fastest yet, but it sure is efficient

Simplify CrossAttention to run on Apple Neural Engine

yeah, giving it a shot now. I want to make sure the einsum I'm using will convert with coremltools though. Will check back in!