Jacob Kahn

@desync-securax — thanks for reporting. I'm still ironing out operators in which we're doing implicit casts (and making these explicit) given some changes to the AMP code. I'm hoping to...

@GabrielLin — do you mean saving the last `n` epochs, or saving _every_ `n` epochs? Saving the last `n` epochs where `n > 1` would have the same effect as...

@tryanbot — can you run the `AllReduceTest` with `mpirun` in the same way you'd start training? Something like ``` mpirun --allow-run-as-root -n 2 ./AllReduceTest train --enable_distributed true --logtostderr=1 ```

@tryanbot — this seems like an issue with your setup or other dependencies. Can you build and run the tests here and see what happens? https://github.com/NVIDIA/nccl-tests

@realdoug — we can definitely take a look. We have several reasons for enforcing having built distributed mode, and we want to use the preprocessor as little as possible, but...

@lunixbochs — we can definitely work towards a smoother build process for Mac. From what I've seen, on MacOS, MKL and Accelerate have essentially the same performance. That said, we...

@lunixbochs — this looks great. I can try to make some of these changes in the next few days. I'm not sure if ArrayFire is [aware of `libmklml`](https://github.com/arrayfire/arrayfire/blob/master/CMakeModules/FindMKL.cmake). As far...

@realdoug — circling back on this after a long time - we recently [released a standalone inference pipeline](https://github.com/facebookresearch/wav2letter/wiki/Inference-Framework) that runs out of the box. The simplest Mac support is still...

@felixt56 — the OpenCL backend isn't supported at the moment. We're working on adding it back in the next several weeks. It should build in theory in v0.3.2 — can...

@datduonguva — if you're building Flashlight libraries and the core only, you can build with `-DFL_LIBRARIES_USE_MKL=OFF` as a CMake option (this isn't documented) which will allow you to use OpenBLAS....