No comments yet.
There's a good bit of info around Tensorflow/Pytorch, not looking for a merits debate (both fantastic imo). I'm with a small AI company supporting both backends in production (linux, bl -> go api interfacing with generic "model" contracts -> calls to libtensorflow.so / libtorch.so via cgo and prototyping in Keras/Pytorch). Small inconvenient though is that it's close to impossible to employ our rag-tag Darwin/GPU/multi-boot local machines to train often as we're stuck with earlier Tensorflow versions and it's quite hard to compile library support locally (Apple doesn't make it easy either) thus often have to rewrite really nice Keras code in Pytorch. Bit of a pity as we have no qualms with Keras, some of our most trusted/understood networks run on it (TF) live.
Anyone happens to know, is there a chance for first-hand support of Tensorflow for OSX in the foreseable future? It has a nice dev ecosystem in general I subjectively believe, for some os us it's a bit like Excel of the finance world :) Totally realize opinions are cheap to offer though, deep gratitude to all people involved in these projects / open source in general, thank you kindly - truly a question, more than glad to accept my ignorance should this sound any other way.
