Support for Mac M1

When building the CPU image on Mac 12.4 Monetary using the M1 architecture tensorflow is throwing an error resulting in the migration scripts to fail so the server is never starting on 8000 and forwarding to 80.

# python manage.py migrate

api_1 | 2022-07-11 18:57:20,770 [nnabla][INFO]: Initializing CPU extension...
api_1 | 2022-07-11 18:58:40.313718: F tensorflow/core/lib/monitoring/sampler.cc:42] Check failed: bucket_limits_[i] > bucket_limits_[i - 1] (0 vs. 10)
api_1 | qemu: uncaught target signal 6 (Aborted) - core dumped
api_1 | Aborted

celery-slow_1  | 2022-07-11 19:00:56,347 [nnabla][INFO]: Initializing CPU extension.. 
celery-slow_1  | 2022-07-11 19:02:07.192204: F tensorflow/core/lib/monitoring/sampler.cc:42] Check failed: bucket_limits_[i] > bucket_limits_[i - 1] (0 vs. 10)
celery-slow_1  | qemu: uncaught target signal 6 (Aborted) - core dumped
celery-slow_1  | ./celery-slow-entrypoint.sh: line 11:    11 Aborted                 celery -A api worker -l INFO -Q slow_queue --concurrency $CELER

Looking at this thread it looks like the issue lies with Tensorflow on Mac M1. A commenter suggest using https://github.com/KumaTea/tensorflow-aarch64 as an alternative to get Tensorflow working with Mac M1.

Hi, can you see if installing this Tensorflow package solves the issue? I got it from the link you provided.

https://snapshots.linaro.org/ldcg/python/tensorflow-manylinux/43/tensorflow-aarch64/tensorflow_aarch64-2.5.3-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl

Did it work?

working on it to see if i can get something working. just built the pre-req naabla for arm64 (download here)

i think we need to build the Dockerfile (backend-api image):

docker build .

Sorry for the very long turnaround, AArch64 support has been added in #930 and is available in the master branch