TensorFlow 2.7 is now released! I tip my hat to all those who helped get this release out 
Release notes: https://github.com/tensorflow/tensorflow/releases/tag/v2.7.0 …
Let me tell you about my favorite feature... 
Show this thread
-
1. Simplified, more focused stack traces. 2. Displaying additional context information in errors that originate from custom Keras layers. 3. A wide-ranging audit of all error messages in Keras and TensorFlow.
Show this thread
About the first one: previously, stack traces would include many frames that originated from TF/Keras internals. They made it harder to find the information you were looking for and to understand what triggered the issue...
-
-
We're now filtering these out, resulting in clean and actionable stack traces. You'll find the root cause faster!pic.twitter.com/pOTtjRUSwu
Show this thread -
About the 2nd one: when an error originates from a custom Keras layer, the 1st thing you do when debugging is to print extensive info about the arguments that were passed to the layer call. We're now providing this context info by default -- as you can see in the image above.
Show this thread -
Lastly: we really believe that the content of error messages is a critical part of the interface of a framework. They should be specific, informative, actionable. We've audited *all* error messages in TF and Keras (literally 1000s of locations) and improved everything we could!
Show this thread -
We hope these changes will make debugging your code faster and more enjoyable, and will generally make your deep learning development experience more productive. Hack on :)
Show this thread
End of conversation
New conversation -
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.