docs(website): add custom transformation to website #116
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #116 +/- ##
=======================================
Coverage 84.54% 84.54%
=======================================
Files 24 24
Lines 1863 1863
=======================================
Hits 1575 1575
Misses 288 288 ☔ View full report in Codecov by Sentry. |
| contents: | ||
| - id: How_to_create_transformer | ||
| title: "How to create your own transformer" | ||
| href: tutorial/How_to_create_your_own_transformer.qmd |
There was a problem hiding this comment.
Use hyphens instead; also, don't capitalize "How"
|
|
||
|
|
||
|
|
||
| ibisML comes with a variety of built-in transformation steps like `OneHotEncode`, `ImputeMean`, `DiscretizeKBins`, and many [others](https://ibis-project.github.io/ibis-ml/reference/steps-outlier.html). However, there are times when you might need to create your own custom preprocessing transformations. This guide will walk you through how to define a custom transformation in ibisML. |
There was a problem hiding this comment.
Why does "others" link to steps-outlier; should it link to https://ibis-project.github.io/ibis-ml/reference/#steps instead?
| from typing import Iterable, Any | ||
| ``` | ||
|
|
||
| ## Implementation Outlines |
There was a problem hiding this comment.
Use sentence case for headings
| Here's how to begin defining the `__init__` method with these considerations: | ||
|
|
||
| ```{python} | ||
| def __init__(self, inputs: SelectionType): |
There was a problem hiding this comment.
Do all these blocks need to get executed? You can use # exec: false annotation (will need to look it up) to make sure these don't get run.
...
Also, when I quickly read through this, I read all of the individual steps, and then I read it again in the final implementation. Is there a better way to present this?
|
|
||
| ```{python} | ||
| # Instantiate CustomRobustScale transformer with the specified columns to scale | ||
| # # Select only one column: "int_col" |
|
|
||
| ### Additional Considerations | ||
|
|
||
| Certainly! Here are additional checks and considerations to ensure the transformer handles unexpected data types or conditions gracefully: |
|
|
||
| Certainly! Here are additional checks and considerations to ensure the transformer handles unexpected data types or conditions gracefully: | ||
|
|
||
| - Check for numeric olumns: Ensure that selected columns are numeric before calculating statistics. This prevents errors when trying to calculate statistics on non-numeric data. |
|
|
||
| - Check for numeric olumns: Ensure that selected columns are numeric before calculating statistics. This prevents errors when trying to calculate statistics on non-numeric data. | ||
| - Check for zero Interquartile Range (IQR): Verify that the IQR (the difference between the 75th and 25th percentiles) is not zero. A zero IQR indicates that all values in the column are the same, making standardization impossible. | ||
| - Backend compatibility: Validate if [operators](https://ibis-project.org/backends/support/matrix) used by ibisML are supported by your chosen backend. This ensures seamless integration and execution of transformations across different environments. |
There was a problem hiding this comment.
Is this a consideration for people creating a step?
| ## Conclusion | ||
|
|
||
| Custom transformers offer a high degree of flexibility and control over data preprocessing tasks. They excel at encapsulating specific steps within the data processing pipeline, which greatly enhances code manageability. If you haven't already, I highly recommend exploring their capabilities and integrating them into your workflow. They can be a valuable asset in streamlining and optimizing your data preprocessing processes. |
There was a problem hiding this comment.
Most FAQs don't have conclusions
|
|
||
| Custom transformers offer a high degree of flexibility and control over data preprocessing tasks. They excel at encapsulating specific steps within the data processing pipeline, which greatly enhances code manageability. If you haven't already, I highly recommend exploring their capabilities and integrating them into your workflow. They can be a valuable asset in streamlining and optimizing your data preprocessing processes. | ||
|
|
||
| ## <span style="color:red">🚀 Contribution Welcome!</span> |
There was a problem hiding this comment.
Why is this randomly red with an emoji
|
|
||
| ## <span style="color:red">🚀 Contribution Welcome!</span> | ||
|
|
||
| Feel free to contribute to our transformations by implementing your own custom transformers or suggesting ones that you find essential. You can do so by checking our transformation [priorities](https://github.com/ibis-project/ibis-ml/issues/32), discussing ideas through creating [issues](https://github.com/ibis-project/ibis-ml/issues), or submitting pull requests (PRs) with your implementations. We welcome collaboration and value input from all contributors. Your ideas and implementations can enrich our library of transformations, making it more comprehensive and useful for everyone involved in data preprocessing tasks. Let's collaborate to enhance the efficiency and effectiveness of our data processing workflows together. |
There was a problem hiding this comment.
Also, docs should usually have a separate page (or pages) on contribution guidelines, not as part of a different FAQ.
It should also include all the information like how to set up your development environment, etc. at some point. This may not be an immediate priority (can check).
|
Can you re-raise this PR, taking into account the feedback, when you get a chance? Temporarily reverted, as getting some more people to view the docs. |
add custom transformation to website
It is part of #69 and #92