Adam Gibson
Adam Gibson
I'll tag @saudet here as well. I might be missing something.
@braeunlich here's what my local run with M1 looks like: ```bash 07:59:49.045 [main] INFO org.nd4j.nativeblas.NativeOpsHolder - Number of threads used for linear algebra: 8 07:59:49.048 [main] INFO org.nd4j.linalg.cpu.nativecpu.CpuNDArrayFactory - Binary...
@braeunlich what happens when you remove cuda? Also, it seems like there's an issue of it loading cuda. It appears to be trying to load cuda 11 and you're loading...
@braeunlich after chatting with @saudet a bit, it does appear to be dll clashes with cpu and cuda. Only one backend can be loaded at a time anyways, so for...
@braeunlich could you DM me on the community forums with a bit of your setup? While I wouldn't be able to offer something for the open source community, there might...
@braeunlich ah before I forget: one thing you could try is setting the cpu priority higher. https://github.com/eclipse/deeplearning4j/blob/master/nd4j/nd4j-backends/nd4j-api-parent/nd4j-api/src/main/java/org/nd4j/linalg/factory/Nd4jBackend.java#L55 This way it doesn't load it if it's not possible. This may work...
@braeunlich yeah we'll document the workaround for now and make sure people are aware of it at a minimum. Thanks for testing!
@MotoCommando we are in the middle of an about finished rewrite of the model import framework. It should be done in a few weeks. For now, the second best option...
@MotoCommando that wasn't really the intended takeaway, Converting a model should only be a few lines of code: https://github.com/onnx/tensorflow-onnx/ the tensorflow import is fairly mature and should suit your needs...
@MotoCommando here's the PR that will address this. https://github.com/eclipse/deeplearning4j/pull/9165 We'll be releasing a release candidate with the new framework soon. This will also replace our existing tensorflow import. I have...