Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ONNX Runtime and CoreML May Silently Convert Your Model to FP16

ym2132.github.io

Post reply on HN
Print view View original

11–19 of 19 posts

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#11

Post by Two_hands »

smcleod wrote: post #10

This was an interesting read, thanks for sharing. I've recently been building something that uses Parakeet v2/v3 models, I'm using the parakeet-rs package ( https://github.com/altunenes/parakeet-rs ) which has had a few issues running models with CoreML (unrelated to the linked post), e.g. https://github.com/microsoft/onnxruntime/issues/26355

Thank you for reading.

Also generally I think CoreML isn't the best. The best solution for ORT would probably be to introduce a pure MPS provider (https://github.com/microsoft/onnxruntime/issues/21271), but given they've already bought into CoreML the effort may not be worth the reward for the core team. Which fair enough as it's a pretty mammoth task

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#13

Post by Two_hands »

yousifa wrote: post #9

On the coreml side this is likely because the neural engine supports fp16 and offloading some/all layers to ANE significantly increases inference time and power usage when running models. You can inspect in the Xcode profiler to see what is running on each part of the device at what precision.

Yeah I can see why they let it be that way, but the fact it is pretty undefined is what bugged me. I suppose it depends on what your goals are - efficiency vs reproducibility.

Also I did run a test of FP16 vs FP32 for a large matmul on the Apple GPU and the FP16 calculation was 1.28x faster so it makes sense that they'd go for FP16 as a default.

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#14

Post by Two_hands »

noosphr wrote: post #5
DiabloD3 wrote: post #2

[flagged]

While this is a bit too harsh - and the solution is naive at best - the problem is real. The idea of bitwise reproducibility for floating point computations is completely laughable in any part of the DL landscape. Meanwhile in just about every other area that uses fp computation it's been the defacto standard for decades. From NVidia not guaranteeing bitwise reproducibility even on the same GPU: https://docs.nvidia.c…

Wow I didn't know that.

The worst part of it is as you say we all accept it and no one talks about it.

Is there any recommended reading you'd suggest to look into this more and the impacts of it?

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#15

Post by noosphr »

Two_hands wrote: post #14
noosphr wrote: post #5

Earlier quoted context omitted.

While this is a bit too harsh - and the solution is naive at best - the problem is real. The idea of bitwise reproducibility for floating point computations is completely laughable in any part of the DL landscape. Meanwhile in just about every other area that uses fp computation it's been the defacto standard for decades. From NVidia not guaranteeing bitwise reproducibility even on the same GPU: https://docs.nvidia.c…

Wow I didn't know that. The worst part of it is as you say we all accept it and no one talks about it. Is there any recommended reading you'd suggest to look into this more and the impacts of it?

Caveat emptor but this seems like an up to date paper on the state of bitwise reproducibility in dl with a bunch of citations to other papers that go into more depth: https://arxiv.org/pdf/2510.09180

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#16

Post by pzo »

Two_hands wrote: post #11
smcleod wrote: post #10

This was an interesting read, thanks for sharing. I've recently been building something that uses Parakeet v2/v3 models, I'm using the parakeet-rs package ( https://github.com/altunenes/parakeet-rs ) which has had a few issues running models with CoreML (unrelated to the linked post), e.g. https://github.com/microsoft/onnxruntime/issues/26355

Thank you for reading. Also generally I think CoreML isn't the best. The best solution for ORT would probably be to introduce a pure MPS provider ( https://github.com/microsoft/onnxruntime/issues/21271 ), but given they've already bought into CoreML the effort may not be worth the reward for the core team. Which fair enough as it's a pretty mammoth task

However one benefits of CoreML - it is the only way to be able for 3rd party to execute on ANE (Apple Neural Engine aka NPU). ANE for some models can execute even faster than GPU/MPS and consume even less battery.

But I agree CoreML in ONNX Runtime is not perfect - most of the time when I tested some models there were too many partitioning and whole graph was running slower compare when using only model in just CoreML format.

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#17

Post by Two_hands »

pzo wrote: post #16
Two_hands wrote: post #11

Earlier quoted context omitted.

Thank you for reading. Also generally I think CoreML isn't the best. The best solution for ORT would probably be to introduce a pure MPS provider ( https://github.com/microsoft/onnxruntime/issues/21271 ), but given they've already bought into CoreML the effort may not be worth the reward for the core team. Which fair enough as it's a pretty mammoth task

However one benefits of CoreML - it is the only way to be able for 3rd party to execute on ANE (Apple Neural Engine aka NPU). ANE for some models can execute even faster than GPU/MPS and consume even less battery. But I agree CoreML in ONNX Runtime is not perfect - most of the time when I tested some models there were too many partitioning and whole graph was running slower compare when using only model in just CoreM…

To be honest it's a shame the whole thing is closed up, I guess it's to be expected from Apple, but I reckon CoreML would be benefit a lot from at least exposing the internals/allowing users to define new ops.

Also, the ANE only allows some operators to be ran on it right? There's very little transparency/control on what can be offloaded to it and cannot which makes using it difficult.

Re: ONNX Runtime and CoreML May Silently Convert Your Model to FP16

#19

Post by Two_hands »

nuc1e0n wrote: post #18

My experiences with ONNX have not been pleasant. Conversions from models written with Tensorflow and Pytorch often fail. I recommend using TFLite or Executorch for deployment to edge devices instead.

Agreed, I have seen some speedups with ONNX if I'm being honest but the process especially on MacOS is a bit messy. I'll try out Executorch and see how it compares, cheers for the recommendation
Post reply on HN