I'm looking to deploy torch models in production and it would be amazing if we could use go for this. As you know containerizing python and pytorch creates a lot of overhead. Not only in terms of memory usage but also the container image is more than a gigabyte in size.
I was looking at https://github.com/orktes/go-torch using torch script but noticed it's out of date and refers to this library. However this library looks more like a complete pytorch replica then for simply deploying models trained in pytorch.
So I'm wondering is this library suitable for deploying torch models? And if so is there some example code? And what fileformat should we use. I believe that pytorch is using pickle to save models, but I'm pretty sure that's not going to work with this go library? There is also torch script but I can't find any reference to it. Or maybe onnx?
I'm looking to deploy torch models in production and it would be amazing if we could use go for this. As you know containerizing python and pytorch creates a lot of overhead. Not only in terms of memory usage but also the container image is more than a gigabyte in size.
I was looking at https://github.com/orktes/go-torch using torch script but noticed it's out of date and refers to this library. However this library looks more like a complete pytorch replica then for simply deploying models trained in pytorch.
So I'm wondering is this library suitable for deploying torch models? And if so is there some example code? And what fileformat should we use. I believe that pytorch is using pickle to save models, but I'm pretty sure that's not going to work with this go library? There is also torch script but I can't find any reference to it. Or maybe onnx?