# Migrate from foundry\_ml \[planned deprecation] to palantir\_models

:::callout{theme="warning" title="Planned deprecation"}
The `foundry_ml` library and dataset-backed models have been fully [deprecated](../../platform-overview/development-life-cycle/index.md) since October 31, 2025, and are unavailable for use. Any remaining workflows should be migrated to use the `palantir_models` library, either by training a new model or by [wrapping the existing model in a model adapter](../migration-reference-faq/index.md#can-a-model-be-migrated-without-re-training). Additionally, models built with `foundry_ml` in Code Workbooks need to be rebuilt in Jupyter® Code Workspaces or Code Repositories. For guidance on building a new model with `palantir_models`, review [how to train a model in Code Repositories](../../model-integration/tutorial-train-code-repositories/index.md) or [how to train a model in Jupyter® notebooks](../../model-integration/tutorial-train-jupyter-notebook/index.md). Contact Palantir Support if you require additional help migrating your workflows.
:::

From version 0.1599.0 onwards, the `palantir_models` library offers [support for Spark ML models](../../integrate-models/spark-models/index.md), but where possible we recommend writing such models to Scikit-learn or a similar single-node framework as part of this migration for a [better experience with Live Inference](../../manage-models/set-up-live/index.md#spark-model-support). We also recommend replacing [MetricSets](../../evaluate-models/metric-sets-reference/index.md) with [experiments](../../model-integration/experiments/index.md) where possible. Experiments now support [image](../../evaluate-models/metric-sets-reference/index.md#image-metrics) metrics, but do not yet support [chart](../../evaluate-models/metric-sets-reference/index.md#chart-metrics) metrics.

## Upgrade campaigns

### Model migration campaign

#### Campaign details

A first campaign, out of two campaigns, was published in [Upgrade Assistant](../../upgrade-assistant/overview/index.md) to help users migrate away from dataset-backed models. Only models that are in use are flagged for review, while others are marked as `Ignored` and filtered out from the campaign view by default.

Users are able to designate a model asset replacement for a dataset-backed model directly from the dataset-backed model page. This information is used by Upgrade Assistant to determine the migration status of the resource. Models with an identified replacement have a status of `Completed` and are filtered out from the campaign view by default. More generally, using this feature is recommended to direct consumers of the model to its replacement in the new framework.

To learn more:

* Review [this example](../how-to/index.md#migration-of-a-model-built-with-foundry_ml-in-code-repositories) to learn how to migrate between the two frameworks;
* Review the [migration FAQ](../migration-reference-faq/index.md) for more details on how the migration works.

#### AIP-powered code suggestions

In environments where [AIP is enabled](../../aip/enable-aip-features/index.md), code migration suggestions powered by a Large Language Model (LLM) through AIP are available. These suggestions can be viewed by selecting the purple icon, as depicted below. The generated code can then be copied using the clipboard icon above each file.

![AIP-powered code suggestions.](../../../assets/a4455921a795ff466a14c93cf29a890476edf8c039592b4ee22a9b622b81b399.webp)

:::callout{theme="warning"}
While the LLM is able to help users to get started with migration, you will likely need to modify the code you are provided by the LLM in order to pass checks and produce a working model. Make sure to thoroughly review the code and the model outputs.
:::

### Consuming resources migration campaign

A second campaign has been published to surface resources which consume deprecated dataset-backed models. These resources include:

* Transforms jobs using dataset-backed models as inputs;
* Code workbooks or transform jobs producing dataset-backed models and other outputs;
* Modeling objectives with a [live](../../manage-models/set-up-live/index.md) or [batch deployment](../../manage-models/set-up-batch/index.md) based on a dataset-backed model.