Release 0.8.8
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.
Additional Functionality
- View Support (Views)
- CloudWatch (CloudWatch Docs)
API Changes
-
Auto Inference name change
When auto_inference is run on an endpoint the name of that inference run is currently
training_holdout
. That is too close tomodel_training
and is confusing. So we're going to change the name toauto_inference
which is way more explanatory and intuitive.Porting plugins: There should really not be any hard coding for
training_holdout
, plugins should just calllist_inference_runs()
(see below) and use the first one on the list. -
list_inference_runs()
The
list_inference_runs()
method on Models has been improved. It now handles error states better (no model, no model training data) and will return 'model_training' LAST on the list, this should improve UX for plugin components.
Examples
model = Model("abalone-regression")
model.list_inference_runs()
Out[1]: ['auto_inference', 'model_training']
model = Model("wine-classification")
model.list_inference_runs()
Out[2]: ['auto_inference', 'training_holdout', 'model_training']
model = Model("aqsol-mol-regression")
model.list_inference_runs()
Out[3]: ['training_holdout', 'model_training']
model = Model("empty-model-group")
model.list_inference_runs()
Out[4]: []
Glue Job Changes
We're spinning up the CloudWatch Handler much earlier now, so if you're setting config like this:
from workbench.utils.config_manager import ConfigManager
# Set the Workbench Config
cm = ConfigManager()
cm.set_config("WORKBENCH_BUCKET", args_dict["workbench-bucket"])
cm.set_config("REDIS_HOST", args_dict["redis-host"])
Just switch out that code for this code. Note: these need to be set before importing workbench
# Set these ENV vars for Workbench
os.environ['WORKBENCH_BUCKET'] = args_dict["workbench-bucket"]
os.environ["REDIS_HOST"] = args_dict["redis-host"]
Misc
Confusion Matrix support for 'ordinal' labels
Pandas has an ‘ordinal’ type, so the confusion matrix method endpoint.confusion_matrix()
now checks the label column to see if it’s ordinal and uses that order, if not just it will alphabetically sort.
Note: This change may not affect your UI experience. Confusion matricies are saved in the Workbench/S3 meta data storage, so a bunch of stuff upstream will also need to happen. FeatureSet object/api for setting the label order, recreation of the model/endpoint and confustion matrix, etc. In general this is a forwarding looking change that will be useful later. :)
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