TensorFlow high-level APIs: Part 3 - Building and refining your models

Welcome to Part 3 of our mini-series on TensorFlow high-level APIs! In this 3 part mini-series, TensorFlow Engineering Manager Karmel Allison runs us through different scenarios using TensorFlow’s high-level APIs.

With the TensorFlow model defined, Karmel discusses how to establish layer architecture, and compile the model, adding the optimizer, loss, and metrics that we are interested in. She also runs through the various steps taken to refine the model.

[davidanalyst]

Every TF video I've ever watched leaves me with "Oh Dang, I've got some more reading and learning to do"... Thanks!!

[wrightw]

Super helpful series. Doesn't just tell you how to build SOME PARTICULAR model, but how to build any basic/canned model. Best mini-tutorial I've seen so far.

[AnthonyBecker]

I really enjoy the presenter's presence & delivery. I feel that it makes the information very easy to take in. Thanks!

[stevenliu]

Thanks a lot. TensorFlow is a powerful, extensible, expressive and beautiful ML platform. I have used it for three years since initial version. This video is very informative.

[clydexu]

concise walk through, something easy to understand and start from.

[tingnews]

That is what that confusion function do. I finally get it. Thank you very much

[bhesht]

Awesome explanation, timely graphics - thank you!

[alexandrumoraru]

TensorFlow is such a great ML Framework. I am so enthusiastic about it :))))

[frigginnobody]

I am trying to use the tf.keras from 2.0, but it is throwing the following error while using LSTMs


"Note that this layer is not optimized for performance. Please use tf.keras.layers.CuDNNLSTM for better performance on GPU"

and the worst part is tf.keras.layers.CuDNNLSTM giving me a name not found!

[SantoshGupta-jnwn]

Great vid! Would love to hear some more details of when to use keras, when to use estimator, when to use graph/session. Is there a difference in how TPU training is done on estimator vs Keras?

[alanzhu]

How do we know what are the optimal number of hidden layer and the neuron within each layer? Is there any module/function in tensorflow or keras that can help?

[ziweiyang]

Do you have any sonnet tutorials for deep complexed model (not just mlp or cnn)?

[BurakBagdatli]

Has anyone figured out how to do the feature layer from the last video? It wasn't working.

[courtney_ann]

How could I prune a model finetuned using a pipeline config file instead of keras

[venushcl]

I am looking for complete code in tensorflow V2.0, could you please share the link. Thanks

[arpitj_]

Where is the link for distribution training?