Recommended workflow for tracking a tuning job on a sklearn model #2453
Unanswered
krishnashanker-amt
asked this question in
Help
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello!
I'm tuning a sklearn model (by providing a training script to sagemaker.sklearn.estimator.SKLearn) with sagemaker.tuner.HyperparameterTuner (mostly similar to the logic followed here ) .
I'm not sure how to track a tuning job setup by this process using Sagemaker Experiments (I know we can track a job in the console). I'm able to link all trials created by this tuning job to an experiment AFTER the tuning job completes using the logic followed here. The
fit()
of HyperparameterTuner does not supportexperiment_config
. I've also tried setting up a tracker in the training script (usingmy_tracker = Tracker.load()
) . This method creates a trial for each training job, butmy_tracker.log_metric(metric_name='test_metric', value=0.4452)
does not work (no metric is logged in the trial component from tracker) .What am I missing? How can I track a tuning job with custom metrics using sagemaker-python-sdk and Sagemaker Experiments?
Beta Was this translation helpful? Give feedback.
All reactions