TensorFlow High-Level APIs: Models in a Box (TensorFlow Dev Summit 2017)

TensorFlow allows you to define models using both low, as well as high-level abstractions. In this talk, Martin Wicke introduces Layers, Estimators, and Canned Estimators for defining models, and shows the roadmap for their availability in core TensorFlow.
Visit the TensorFlow website for all session recordings: goo.gl/bsYmza
Subscribe to the Google Developers channel at goo.gl/mQyv5L
event: TensorFlow Dev Summit 2017; re_ty: Publish;

[@MartinAndrews-mdda]

Should the 'classifier.fit()' and 'classifier.evaluate()' calls at ~12m50s should be on 'regressor'?

[@sethguythewizard]

how can i get one of those orange shirts with the node graph

[@ashpats2]

Black jumper with node icon :)

[@Mrbooboo1972]

Awesome. Perhaps you could add a bit more Volume when doing these videos. Please. Have my volume maxed and it's difficult to hear him.

[@jessiechan7629]

lol maybe you just need a new laptop, I'm hearing fine on mine.

[@Qwantopides]

It is a little quiet to me too.

[@SantoshGupta-jn1wn]

It seems that the sessions run through a Estimator are not quite the same as a normal session? For example, `.eval.()` operation and the `tf.nn.sampled_softmax_loss` function do not work, when they do for a normal session. Would love to hear some details on this phenomenon.

[@redberries8039]

I assume i can define a model using layers then also use an estimator ....but the estimator seemed to include a model definition ..... so can the model inside that estimator box be the model defined with the layers? I guess it can be, as that would make sense, but I feel uncertain what's going on.

[@redberries8039]

ummm I'm thinking now estimators define their own model and layers are part of them [maybe internally, but not as far as we are concerned] ??

[@Mrbooboo1972]

Interesting