Query model metadata#

You can download model metadata from an individual run, or fetch model metadata from all runs in the project as a data frame.

Fetching from single run#

When operating on one run at a time, you can fetch and download any model artifacts and other metadata logged to the run object.

Initialize the related run:

import neptune

run = neptune.init_run(with_id="CLS-56", mode="read-only")

How do I find the ID?

The Neptune ID is a unique identifier for the run. The Experiments tab displays it in the leftmost column.

In the run structure, the ID is stored in the system namespace (sys).

If the run is active, you can obtain its ID with run["sys/id"].fetch(). For example:
```
>>> run = neptune.init_run()
...
>>> run["sys/id"].fetch()
'CLS-26'
```

If you set a custom run ID, it's stored at sys/custom_run_id:

>>> run["sys/custom_run_id"].fetch()
'vigilant-puffin-20bt9'

Example: Downloading a model signature, if it exists

if run.exists("model/signature"):
    run["model/signature"].download()

For more examples, see How to query metadata via the Neptune API.

Fetching from multiple runs as data frame#

To fetch multiple runs and filter them by model metadata, pass an NQL string to the query argument of the fetch_runs_table() function.

The below example assumes the following:

The stage is logged to the string field model/stage. For example, run["model/stage"] = "production".
The model size is logged to the float field model/size. For example, run["model/size"] = "10256.0".
A test accuracy score is logged to the field test/acc. For example, run["test/acc"] = 0.87.

The snippet fetches runs with:

model versions in production,
the model stage, size, and test accuracy as columns.

It also sorts the entries by model size.

Example: Fetch runs with models in production

project = neptune.init_project(
    project="workspace-name/project-name", # (1)! 
    mode="read-only",
)

models_df = project.fetch_runs_table(
    query='`model/stage`:string = "production"',
    columns=["model/stage", "model/size", "test/acc"],
    sort_by="model/size",
).to_pandas()

The full project name. For example, "ml-team/classification".
- You can copy the name from the project details ( → Details & privacy)
- You can also find a pre-filled project string in Experiments → Create a new run.

For the syntax and examples, see the Neptune Query Language (NQL) reference.