Research with TensorFlow (TF Dev Summit '20)

In this talk we'll go over some interesting features of TF which are useful when doing research.

Speaker:
Alexandre Passos - Software Engineer

event: TensorFlow Dev Summit 2020; re_ty: Publish; product: TensorFlow - TensorFlow Research Cloud; fullname: Alexandre Passos;

[GauravSharma-uiyd]

One thing I noticed that whenever I do something in tf or watch videos or read blogs on tf I progressively get lost in the syntax and programmatic jargons and heavy lifting and forget my primary purpose of using it that is Deep learning. It seems like we should first do a course on how to read, write and interpret tf and then do ML and DL.
It's not friendly for beginners like pytorch.
But keras is awesome :)

[ravivijayk]

thanks Alexandre Passos for walkthrough
watched some of the sections at least 4-5 times by now, planning to execute example code and post if any questions. #RaggedTensor

[rameezismail]

Thank you for sharing, I really like TensorFlow's Core API. keep it up!
May be not a popular opinion, but I think once you have got good grasp of the basic concepts, the versatility and power of data flow modeling become apparent. I find keras too much abstracted, not a very good option for me.

[cartoonchan]

I tried to learn tensorflow but code for tf and its concept like variables are just way to tough for me. Can anyone, who reads this comment tell me how did you understand tf and how did you learned the coding. It'll be get if you post some link. Thank you