Applicable to: Customers using the AI Deployment product — experiment tracking, job runs, and ML repositories.If you use only AI Gateway and do not deploy jobs or log experiment runs, this change does not affect you.
This change affects you if your job-run or training scripts call run.log_images() or run.log_plots(), or you view logged images and plots in the Images and Plots tabs on a run’s detail page.This change does not affect you if you log image or plot files with run.log_artifact() — those continue to work as before and appear under the Artifacts tab.
What Is Changing
Starting with v0.156.0, dedicated support for logging and visualizing images and plots on experiment runs is removed.
run.log_images() and run.log_plots() are no longer supported for new runs.- The dedicated Images and Plots tabs on run detail pages are removed.
- All existing logged images and plots are automatically migrated to standard artifacts when you upgrade to v0.156.0. Migrated files appear under the Artifacts tab on the associated run.
This is a breaking change for any workflow that relies on the dedicated image and plot logging APIs or the in-dashboard visualization for those artifact types.
Why This Change
Image and plot logging sees very limited adoption across job runs. For the typical job-run workflow — persisting training outputs, evaluation figures, or diagnostic files — standard artifacts already cover the same use case:
- Save image or plot files to disk during your job.
- Log them with
run.log_artifact() and ArtifactPath — pass step the same way you did with log_images() and log_plots().
- Access them from the Artifacts tab on the run, or download them via the SDK.
This simplifies the platform surface area without losing the ability to capture visual outputs from your jobs.
What You Need to Do
Upgrade to v0.156.0
On upgrade, existing images and plots logged on prior runs are migrated to artifacts automatically. No manual data migration is required.
Update your scripts
Search your training scripts, notebooks, and job definitions for log_images, log_plots, and any references to image or plot artifact types. Replace them with run.log_artifact() before your next job run on v0.156.0 (see examples below).
Validate your job runs
Run a test job and confirm that logged files appear under the Artifacts tab on the run detail page.
Migrating plots
Before:
import matplotlib.pyplot as plt
from sklearn.metrics import ConfusionMatrixDisplay
ConfusionMatrixDisplay.from_predictions(["spam", "ham"], ["ham", "ham"])
run.log_plots({"confusion_matrix": plt}, step=1)
import matplotlib.pyplot as plt
from sklearn.metrics import ConfusionMatrixDisplay
from truefoundry.ml import ArtifactPath
ConfusionMatrixDisplay.from_predictions(["spam", "ham"], ["ham", "ham"])
plt.savefig("confusion_matrix.png")
run.log_artifact(
name="evaluation-plots",
artifact_paths=[ArtifactPath(src="confusion_matrix.png")],
step=1,
)
fig.write_html("distribution-plot.html")
run.log_artifact(
name="evaluation-plots",
artifact_paths=[ArtifactPath(src="distribution-plot.html")],
step=1,
)
Migrating images
Before:
from truefoundry.ml import Image
images_to_log = {
"logged-image-from-path": Image(data_or_path="result_image.jpeg"),
}
run.log_images(images_to_log, step=1)
from truefoundry.ml import ArtifactPath
run.log_artifact(
name="training-images",
artifact_paths=[ArtifactPath(src="result_image.jpeg")],
step=1,
)
log_artifact reference.
Update any scripts that call log_images or log_plots before running new jobs on v0.156.0.
If you have questions or need help migrating image and plot logging to artifacts, reach out to support@truefoundry.com — we’re happy to assist.
