Closed asheetal closed 3 years ago
Hi, thank you for filing. This example is indeed broken.
Some investigative notes:
This is the PR where the function was removed: https://github.com/rstudio/tfdatasets/pull/67
At a later time, the corresponding functions in the tfhub R package were refactored and renamed to step_pretrained_text_embedding
and step_pretrained_text_embedding
. However, those functions after refactoring are not compatible with the tfdatasets steps_
+ feature_spec()
family of functions, but are geared towards the recipes
steps in the tidymodels ecosystem. Fixing step_pretrained_text_embedding
to work with tfdatasets feature_spec()
is going to be a large effort and will likely not happen.
The current design of the feature_spec()
approach is geared towards tfestimators. However, tfestimators is deprecated, and all the upstream effort is on keras
+ keras.preprocessing
.
Resolution: removed the broken example.
This tutorial uses function call to step_text_embedding_column using pretrained tfhub module_spec https://tensorflow.rstudio.com/guide/tfhub/examples/feature_column/
I am using the tensorflow 2.4 and latest packages. But the tutorial does not run because the function does not exist in latest tfdatasets. Can someone point me to the right direction?