Skip to main content

Fixed: "RESOURCE_DOES_NOT_EXIST: Registered Model with name=m-… not found" When You Download an MLflow Model File

· 5 min read
Jagdish Kumawat
Founder @ Dewiride

I wanted one file from a logged model, its serving_input_example.json, and asked MLflow for models:/<model ID>/serving_input_example.json. It answered RESOURCE_DOES_NOT_EXIST: Registered Model with name=m-6e2a8bb07fc44b02843eca07c555586f not found. A models:/ URI with a path after it is read as a registered model's name and version, not as a logged model's ID and a file. The model's own artifact location, an mlflow-artifacts:/ URI, plus the file name, downloaded just that file.

What I Was Doing​

For Part 6 of the MLflow series, I wanted to test the shopper model with MLflow's command line, using the request the model was logged with, serving_input_example.json. On MLflow 3.18.0, in the project folder, I asked for that one file:

Terminal
uv run mlflow --env-file .env artifacts download -u models:/m-6e2a8bb07fc44b02843eca07c555586f/serving_input_example.json -d downloaded-file

What I Expected​

downloaded-file/serving_input_example.json, and the path printed.

What I Got​

Output (first and last lines)
Loaded environment variables from: .env
Traceback (most recent call last):
...
mlflow.exceptions.RestException: RESOURCE_DOES_NOT_EXIST: Registered Model with name=m-6e2a8bb07fc44b02843eca07c555586f not found

The download command in a terminal: Loaded environment variables from .env, a traceback, and at the end mlflow.exceptions.RestException: RESOURCE_DOES_NOT_EXIST: Registered Model with name=m-6e2a8bb07fc44b02843eca07c555586f not found.

A registered model? This model was never registered; it's a logged model, and models:/m-6e2a8bb07fc44b02843eca07c555586f alone, without the file name, downloads it fine.

What I Checked​

The URIs MLflow accepts. The reference for mlflow.artifacts.download_artifacts lists the formats, among them models:/<model_name>/<stage>, models:/<model_name>/<version>/path/to/model and models:/<model_name>@<alias>/path/to/model: every models:/ URI with a path after it starts with a registered model's name. A path after a logged model's ID isn't among them.

Where the model's files are. A logged model has an artifact location of its own, which mlflow.get_logged_model returns:

Terminal
uv run python -c "import mlflow; mlflow.set_tracking_uri('http://127.0.0.1:8080'); print(mlflow.get_logged_model('m-6e2a8bb07fc44b02843eca07c555586f').artifact_location)"
Output
mlflow-artifacts:/3/models/m-6e2a8bb07fc44b02843eca07c555586f/artifacts

The model's Artifacts tab in the UI shows the same address, as Path:, above each file.

Why It Happens​

models:/a/b is the registered-model form: model a, version b. So MLflow took m-6e2a8bb0… for a registered model's name and serving_input_example.json for a version, asked the registry for that model, and the registry had no model by that name.

The Fix​

Download from the model's artifact location, with the file name after it:

Terminal
uv run mlflow --env-file .env artifacts download -u mlflow-artifacts:/3/models/m-6e2a8bb07fc44b02843eca07c555586f/artifacts/serving_input_example.json -d downloaded-file

After the Fix​

Output
Loaded environment variables from: .env

/Users/you/source/dewiride-shopper-purchase-prediction/downloaded-file/serving_input_example.json

The fixed command's output in a terminal: Loaded environment variables from .env, then the path of the downloaded serving_input_example.json in the downloaded-file folder.

Just that file, in downloaded-file. To fetch the whole model instead, models:/<model ID> without a path works, as How to Download an MLflow Model's Files shows.

Common Mistakes​

MistakeWhat happensDo this instead
models:/<model ID>/<file>Read as a registered model and version: Registered Model with name=m-... not found<artifact location>/<file>
Building the location by hand from the experiment's IDEasy to get wrongCopy it from get_logged_model(...).artifact_location or the UI's Path:

Troubleshooting​

SymptomLikely causeFix
RESOURCE_DOES_NOT_EXIST: Registered Model with name=m-... not foundA path after a logged model's ID in a models:/ URIUse the artifact location
Logged model with ID '...' not found.MLFLOW_TRACKING_URI isn't set, so MLflow looks in a local databaseSet it, or pass --env-file .env as above

FAQ​

Do registered models have the same problem?​

No: models:/<name>/<version>/path is the documented form for them. The trouble is only a path after a logged model's m-… ID.

Can I download a file with Python instead?​

Yes: mlflow.artifacts.download_artifacts(artifact_uri="mlflow-artifacts:/.../artifacts/serving_input_example.json", dst_path="downloaded-file") takes the same URI.

Conclusion​

models:/<model ID>/serving_input_example.json failed because a models:/ URI with a path is a registered model's address. The logged model's artifact location, mlflow-artifacts:/3/models/m-…/artifacts, plus the file name, downloaded the one file I wanted.

Additional Resources​