rip all my existing models
-
-
- End of conversation
New conversation -
-
-
I'm not. I'm disappointed that TF2.0 managed to completely change most of the API – breaking all of the old TF tutorials (>80%). I tried porting some TF1.0 code to TPU. You're supposed to use TPUEstimator, but Google doesn't even have an example notebook that uses it.
-
Of the code that does still work, most of it will need to be rewritten. It's very hard to find a TF codebase that doesn't generate compatibility warnings, even if it's a TF2.0 sample. I realize that things will get better with time, but it's eerily reminiscent of Python2 -> 3.
- Show replies
New conversation -
-
-
This Tweet is unavailable.
-
Having to rewrite your entire codebase to use the new API is pretty exciting. It will also be especially challenging, since there are almost no tutorials for TF2.0. Everyone loves a good challenge.
End of conversation
-
-
-
And you get a deprecation warning! And YOU get a deprecation warning! Everyone gets a deprecation warning!pic.twitter.com/lII036T35F
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Looking forward to being able to save and load models built with the functional and DenseFeatures api.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
Looking forward to see a new face for an enormous community driven project, full of deprecations, warnings, duplicated submodules and poor documentation. New major version, new life.
Thanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.