MLBazaar / MLBazaar/MLBlocks

MLPipeline does not preserve metadata from JSON pipeline annotation

Open
#137 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
124
Forks
33
PR merge metrics
No merged PRs in 30d

Description

  • MLBlocks version: 0.4
  • Python version: 3.8
Description

I'm trying to get metadata from an MLPipeline object that was present in the JSON pipeline annotation that was loaded. For example, the annotation has a metadata.name field that I'd like to access from the pipeline.

What I Did

In the following example, I would expect that the MLPipeline has a metadata dict which has name key, just like the JSON. But it doesn't.

$ mkdir -p mlprimitives mlpipelines
$ curl -s https://raw.githubusercontent.com/MLBazaar/MLPrimitives/master/mlprimitives/primitives/sklearn.ensemble.RandomForestRegressor.json -o mlprimitives/sklearn.ensemble.RandomForestRegressor.json
$ curl -s https://raw.githubusercontent.com/MLBazaar/MLPrimitives/master/mlprimitives/pipelines/sklearn.ensemble.RandomForestRegressor.json -o mlpipelines/sklearn.ensemble.RandomForestRegressor.json
$ jq .metadata.name mlpipelines/sklearn.ensemble.RandomForestRegressor.json 
"sklearn.ensemble.RandomForestRegressor"
$ python
Python 3.8.3 (default, Jul 20 2020, 16:43:14) 
[Clang 11.0.3 (clang-1103.0.32.62)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from mlblocks import load_pipeline, MLPipeline
>>> pipeline = MLPipeline(load_pipeline('sklearn.ensemble.RandomForestRegressor'))
>>> pipeline.metadata
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'MLPipeline' object has no attribute 'metadata'
Suggestions
  1. Explicitly support persisting metadata on the MLPipeline object (and presumably on underlying MLBlock objects)
  2. Raise an error of the JSON input contains unused (unsupported) keys
  3. Guarantee that MLPipeline.load and MLPipeline.save are inverse operations, i.e. that no data is lost (currently metadata fields are lost)

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with the MLPipeline and load_pipeline entry points, then trace how the JSON pipeline annotation is loaded and how MLPipeline.load and MLPipeline.save handle data. Compare the annotation's metadata.name with the resulting object and saved representation; done requires an agreed behavior for preserving metadata or rejecting unsupported keys.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.