Skip to main content

Fixed: UserWarning "The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource" in MLflow

· 10 min read
Jagdish Kumawat
Founder @ Dewiride

Logging a pandas DataFrame read from a CSV file, with that file as its source, made MLflow print UserWarning: The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource. It names the same source type twice because MLflow 3.18.0 registers it twice. The dataset it logs is correct, and a warnings filter for this one message keeps your output clean until MLflow fixes it.

What I Was Doing​

I was writing How to Log a Dataset in MLflow, with MLflow 3.18.0, pandas 3.1.1 and Python 3.15.2. Its example reads a small CSV file, houses.csv, and records it as an MLflow dataset whose source is that file:

log_dataset.py
import mlflow
import pandas as pd

mlflow.set_tracking_uri("http://127.0.0.1:8080")
mlflow.set_experiment("datasets-demo")

houses = pd.read_csv("houses.csv")
dataset = mlflow.data.from_pandas(houses, source="houses.csv", targets="price", name="houses")

print("Name: ", dataset.name)
print("Digest: ", dataset.digest)
print("Profile:", dataset.profile)
print("Schema: ", dataset.schema)

with mlflow.start_run(run_name="log-a-dataset"):
mlflow.log_input(dataset, context="training")
Terminal
uv run python log_dataset.py

What I Expected​

The four lines the script prints, then MLflow's links to the run.

What I Got​

Output
2026/11/19 16:21:35 INFO mlflow.tracking.fluent: Experiment with name 'datasets-demo' does not exist. Creating a new experiment.
/Users/you/source/mlflow-demo/.venv/lib/python3.15/site-packages/mlflow/data/dataset_source_registry.py:148: UserWarning: The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource. MLflow will assume that this is a LocalArtifactDatasetSource source.
return _dataset_source_registry.resolve(
Name: houses
Digest: 24586459
Profile: {'num_rows': 5, 'num_elements': 15}
/Users/you/source/mlflow-demo/.venv/lib/python3.15/site-packages/mlflow/types/utils.py:440: UserWarning: Hint: Inferred schema contains integer column(s). Integer columns in Python cannot represent missing values. If your input data contains missing values at inference time, it will be encoded as floats and will cause a schema enforcement error. The best way to avoid this problem is to infer the model schema based on a realistic data sample (training dataset) that includes missing values. Alternatively, you can declare integer columns as doubles (float64) whenever these columns may have missing values. See `Handling Integers With Missing Values <https://www.mlflow.org/docs/latest/models.html#handling-integers-with-missing-values>`_ for more details.
warnings.warn(
Schema: ['rooms': long (required), 'size_m2': long (required), 'price': long (required)]
🏃 View run log-a-dataset at: http://127.0.0.1:8080/#/experiments/6/runs/257be85c3a7447228d0b2bcab91cd0e0
🧪 View experiment at: http://127.0.0.1:8080/#/experiments/6

A terminal running uv run python log_dataset.py: after the experiment is created, a UserWarning from dataset_source_registry.py line 148 says the specified dataset source can be interpreted in multiple ways, LocalArtifactDatasetSource, LocalArtifactDatasetSource, then the dataset's name, digest and profile, a second warning about integer columns, the schema and the run links.

Two warnings. This post is about the first, which appeared when from_pandas() read the source argument. The second, about integer columns, has its own post: Fixed: UserWarning "Hint: Inferred schema contains integer column(s)".

What I Checked​

What raises it. Line 148 of mlflow/data/dataset_source_registry.py is where from_pandas() asks MLflow's list of dataset source types which one fits "houses.csv". In MLflow 3.18.0, the registry warns whenever more than one type says yes, and lists each one's class name:

mlflow/data/dataset_source_registry.py (MLflow 3.18.0)
if len(matching_sources) > 1:
source_class_names_str = ", ".join([source.__name__ for source in matching_sources])
warnings.warn(
f"The specified dataset source can be interpreted in multiple ways:"
f" {source_class_names_str}. MLflow will assume that this is a"
f" {matching_sources[-1].__name__} source.",
stacklevel=2,
)

Why the same name twice. MLflow builds a dataset source type for each kind of artifact storage it knows, keyed by the start of an address, such as s3 or file. Two of those keys lead to the same local storage, LocalArtifactRepository: an empty key for plain paths and file for file:// addresses. I asked MLflow, through two of its internal modules, which keys use local storage and which source types accept houses.csv:

Python (a check using MLflow's internal modules)
from mlflow.data.dataset_source_registry import _dataset_source_registry as reg
from mlflow.store.artifact.artifact_repository_registry import get_registered_artifact_repositories
repos = get_registered_artifact_repositories()
print("schemes using LocalArtifactRepository:", [repr(s) for s, r in repos.items() if r.__name__ == "LocalArtifactRepository"])
matches = [s for s in reg.sources if s._can_resolve("houses.csv")]
print("sources that can resolve houses.csv:", [(s.__name__, s._get_source_type()) for s in matches])
Output
schemes using LocalArtifactRepository: ["''", "'file'"]
sources that can resolve houses.csv: [('LocalArtifactDatasetSource', 'local'), ('LocalArtifactDatasetSource', 'local')]

So there are two separate source types, both named LocalArtifactDatasetSource, both of kind local, and each accepts any local path.

Whether the logged dataset is right. Either type records the same thing: kind local and the path as given. The run's dataset panel shows exactly that:

The Data details panel of the log-a-dataset run in the MLflow UI: one dataset used, houses (24586459) with the Training context, 5 rows, 15 elements, digest 24586459, source type Local and path houses.csv, and a schema of three long columns: rooms, size_m2 and price.

Other ways to write the path. A pathlib.Path object, a relative path such as data/houses.csv and a full file:///… address all produced the same warning.

MLflow's issue tracker. The bug is mlflow/mlflow#9702, "[BUG] Warning about LocalArtifactDatasetSource being ambiguous with itself", open since September 22, 2023, when it was reported against MLflow 2.7.1. Its reporter wrote: "I can't tell that there's anything I can do in my application code to fix this." A pull request to register local storage once, #26363, was closed without being merged on October 14, 2026, with the comment "Closing because #9702 is over a year old and needs to be re-triaged before we take a fix."

MLflow's documentation. from_pandas() documents source as "a filesystem path, an S3 URI, an HTTPS URL" and more, "or an instance of DatasetSource". The local source types are built while MLflow loads and aren't part of its API reference, so there is no documented way to hand one over and skip the lookup.

Why It Happens​

MLflow 3.18.0 registers its local storage under two keys, an empty one and file, and builds a dataset source type for each. Both answer yes for any local path, so the lookup in from_pandas() finds two matches with the same name and warns. It then uses the last match, which records the path exactly as the first would.

The Fix​

Tell Python to ignore this one warning before the dataset is created. Add the filter at the top of the script:

log_dataset.py (the top of the file)
import warnings

import mlflow
import pandas as pd

# MLflow 3.18 registers local files under two identical source types and warns about
# it (mlflow/mlflow#9702). Both give the same record, so ignore just this warning.
warnings.filterwarnings(
"ignore",
message="The specified dataset source can be interpreted in multiple ways",
category=UserWarning,
)

The top of log_dataset.py in a macOS-style code window: imports of warnings, mlflow and pandas, a comment that MLflow 3.18 registers local files under two identical source types (mlflow/mlflow#9702), and a warnings.filterwarnings call that ignores UserWarning messages starting with The specified dataset source can be interpreted in multiple ways.

  • warnings.filterwarnings("ignore", ...) adds a rule to Python's warning filters: matching warnings aren't shown.
  • message is a regular expression that the start of the warning must match, so the rule catches only this MLflow message.
  • category=UserWarning is the warning's class, from the output above.
  • The comment names the issue, so the filter can go once a fixed MLflow release is out.

After the Fix​

Terminal
uv run python log_dataset.py
Output
Name: houses
Digest: 24586459
Profile: {'num_rows': 5, 'num_elements': 15}
/Users/you/source/mlflow-demo/.venv/lib/python3.15/site-packages/mlflow/types/utils.py:440: UserWarning: Hint: Inferred schema contains integer column(s). Integer columns in Python cannot represent missing values. If your input data contains missing values at inference time, it will be encoded as floats and will cause a schema enforcement error. The best way to avoid this problem is to infer the model schema based on a realistic data sample (training dataset) that includes missing values. Alternatively, you can declare integer columns as doubles (float64) whenever these columns may have missing values. See `Handling Integers With Missing Values <https://www.mlflow.org/docs/latest/models.html#handling-integers-with-missing-values>`_ for more details.
warnings.warn(
Schema: ['rooms': long (required), 'size_m2': long (required), 'price': long (required)]
🏃 View run log-a-dataset at: http://127.0.0.1:8080/#/experiments/6/runs/d6565b948bca424e941ba7d30e782bf9
🧪 View experiment at: http://127.0.0.1:8080/#/experiments/6

A terminal running uv run python log_dataset.py after the fix: the dataset's name, digest 24586459 and profile print without the dataset source warning; only the separate integer-column hint remains before the schema and the run links.

The source warning is gone, and the dataset is unchanged: same digest, 24586459. The experiment already existed, so MLflow no longer says it's creating one. The integer hint that remains has its own fix.

Common Mistakes​

MistakeWhat happensDo this instead
Adding the filter after from_pandas()The warning has already been shownAdd it before the first from_pandas()
warnings.simplefilter("ignore")Every warning disappears, including ones you needFilter by message and category
Dropping source to make the warning goThe run records your script as the source instead of the fileKeep source and filter the warning

Troubleshooting​

SymptomLikely causeFix
The warning still showsThe filter runs after the dataset is created, or the message has a typoPut the filter at the top and copy the message's start exactly
MlflowException: Could not find a source information resolver for the specified dataset source: 42.source got something that is neither a path nor an address, here a numberPass a path, a file:// address or a URL such as https://…

FAQ​

Is the dataset logged differently because of the warning?​

No. Both matching types are local sources and store the path as you gave it. The run shows Source type: Local and Path: houses.csv either way.

Does an HTTPS source warn too?​

No. Only local storage is registered twice. A dataset whose source is an https:// address resolves to a single type, HTTPDatasetSource, without a warning.

When can I delete the filter?​

When an MLflow release closes issue #9702. Run your script once without the filter after upgrading, and delete it if the warning is gone.

Conclusion​

The specified dataset source can be interpreted in multiple ways: LocalArtifactDatasetSource, LocalArtifactDatasetSource comes from MLflow 3.18.0 registering local storage under two keys, so a local path matches two identical source types. The dataset is recorded correctly either way. A warnings.filterwarnings rule for that one message, placed before the first from_pandas(), hides it until MLflow fixes issue #9702.

Additional Resources​