Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Fix the table view issue in data transform design doc. #1538

Merged
merged 2 commits into from Nov 29, 2019
Merged

Fix the table view issue in data transform design doc. #1538

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
877fa64
Fix the format issue for the table view
brightcoder01 Nov 29, 2019
2adeba4
Update the issue link
brightcoder01 Nov 29, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
5 changes: 3 additions & 2 deletions docs/designs/data_transform.md
View file
Open in desktop
Expand Up @@ -41,7 +41,8 @@ In the Analyze step, we will parse the TRANSFORM expression and collect the stat
In the feature column generation step, we will format the feature column template with the variable name and the statistical values to get the integral feature column definition for the transform logic.
The generated feature column definitions will be passed to the next couler step: model training. We combine them with the COLUMN expression to generated the final feature column definitions and then pass to the model. Let's take **NUMERIC(STANDARDIZE(age))** for example, the final definition will be **numeric_column('age', normalizer_fn=lambda x: x - 18.0 / 6.0)**

We plan to implement the following common used transform APIs at the first step. And we will add more according to further requirements.
We plan to implement the following common used transform APIs at the first step. And we will add more according to further requirements.

| Name | Feature Column Template | Analyzer |
|:---------------------------:|:------------------------------------------------------------------------------:|:------------------:|
| STANDARDIZE(x) | numeric_column({var_name}, normalizer_fn=lambda x : x - {mean} / {std}) | MEAN, STDDEV |
Expand All @@ -60,5 +61,5 @@ This solution can bring the following benifits:

We need figure out the following points for this further solution:

1. Model Export: Upgrade keras API to support exporting the transform logic and the model definition together to SavedModel for inference. [Issue](https://github.com/tensorflow/transform/issues/150)
1. Model Export: Upgrade keras API to support exporting the transform logic and the model definition together to SavedModel for inference. [Issue](https://github.com/tensorflow/tensorflow/issues/34618)
2. Transform Execution: We will transform the data records one by one using the transform logic in the SavedModel format and then write to a new table. We also need write a Jar, it packages the TensorFlow library, loads the SavedModel into memory and processes the input data. And then we register it as UDF in Hive or MaxCompute and use it to transform the data.