Release 0.8.55

Need Help?

The SuperCowPowers team is happy to give any assistance needed when setting up AWS and Workbench. So please contact us at workbench@supercowpowers.com or on chat us up on Discord

The Workbench framework continues to flex to support different real world use cases when operating a set of production machine learning pipelines.

Note: These release notes cover the changes from 0.8.50 to 0.8.55

General

This release is an incremental release as part of the road map for v.0.9.0. Please see the full details of the planned changes here: v0.9.0 Roadmap.

FeatureSet: training column values

We're got a good suggestion from one of our beta customers to change the training column to use True/False values instead of 1/0. Having boolean values make semantic sense and make filtering easier and more intuitive.

API Changes

FeatureSet Queries

Since the training column now contains True/False, any code that you have where you're doing a query against the training view.

fs.query(f'SELECT * FROM "{table}" where training = 1')
<changed to>
fs.query(f'SELECT * FROM "{table}" where training = TRUE')

fs.query(f'SELECT * FROM "{table}" where training = 0')
<changed to>
fs.query(f'SELECT * FROM "{table}" where training = FALSE')

Also dataframe filtering is easier now, so if you have a call to filter the dataframe that also needs to change.

df_train = all_df[all_df["training"] == 1].copy()
<changed to>
df_train = all_df[all_df["training"]].copy()

df_val = all_df[all_df["training"] == 0].copy()
<changed to>
df_val = all_df[~all_df["training"]].copy()

For more details see: Training View

Model Instantiation

We got a request to reduce the time for Model() object instantiation. So we created a new CachedModel() class that is much faster to instantiate.

%time Model("abalone-regression")
CPU times: user 227 ms, sys: 19.5 ms, total: 246 ms
Wall time: 2.97 s

%time CachedModel("abalone-regression")
CPU times: user 8.83 ms, sys: 2.64 ms, total: 11.5 ms
Wall time: 22.7 ms

For more details see: CachedModel

Improvements

Workbench REPL Onboarding

At some point the onboarding with Workbench REPL got broken and wasn't properly responding when the user didn't have a complete AWS/Workbench setup.

Internal Changes

The decorator for the CachedMeta class did not work properly in Python 3.9 so had to be slightly refactored.

Specific Code Changes

Code Diff v0.8.50 --> v0.8.55

Who doesn't like looking at code! Also +3 points for getting down this far! Here's a cow joke as a reward:

That feeling like you’ve done this before? .... Deja-moo

Questions?

The SuperCowPowers team is happy to answer any questions you may have about AWS and Workbench. Please contact us at workbench@supercowpowers.com or on chat us up on Discord