df-ml-anomaly-detection icon indicating copy to clipboard operation
df-ml-anomaly-detection copied to clipboard

Breaking Change with TF 2.0

Open santhh opened this issue 4 years ago • 2 comments

We currently use model = tf.contrib.estimator.forward_features(model,KEY_COLUMN) for the model creation which not supported for TF 2.0. Need to upgrade the code for TF 2.0 https://www.tensorflow.org/guide/upgrade @GoogleCloudPlatform/streaming-analytics-ai

santhh avatar Aug 20 '20 20:08 santhh

I wrote up a blog post and example notebook on how to do this in TF 2.x:

https://towardsdatascience.com/how-to-extend-a-keras-model-5effc083265c https://github.com/GoogleCloudPlatform/training-data-analyst/blob/master/blogs/batch_predictions/batch_predictions_keras.ipynb

dhodun avatar Aug 25 '20 14:08 dhodun

thanks @dhodun I will look. If you have some bandwidth-please feel free to contribute too.

santhh avatar Aug 27 '20 19:08 santhh