Any HNers doing deep learning with TF?<p>I am currently learning Tensorflow, and seems the API is too big, learning curve is too steep compared to Theano, Pytorch and Chainer.<p>I can't get doing declarative programming in Python and Tensorflow. Lots of scoping with with (...) :<p>Even Theano is more intuitive (at least for me) and make more sense in its API, even though it also has lots of hacks and is define-then-run like TF.<p>How do you get good in using Tensorflow, create models, using tfdbg, how to step trace your code, using withs, etc. ?<p>Thanks.
The big pieces that have helped me understand and grow with tf:<p>I began learning a lot about the workflow through cookbooks with dialogue on why specific practices are used. My favorites started small with something like a linear regression model and hit all the major notes up to neural nets. If you go this route, I would pick something published recently so you are starting with the current version and can learn the new features as they are released.<p>From there I could begin to understand online tutorials on implementing this or that model. This gave me a lot of context on how I could begin testing different ideas and tuning my hyper parameters in my own models.<p>Someone who is a true pro can take if from here.
Disclaimer: I write competing software to tensorflow and compete with google cloud which commercializes tensorflow.<p>So let me turn this on its head. Why does it matter? Use what you find works for you. If you already know the concepts, what's stopping you from just using pytorch for now?<p>I would suggest stepping back and asking yourself that first. Don't learn something "cuz hype".<p>Granted, TF has the new eager api that just came out. That might help you?