Skip to content

Commit

Permalink
model registry guide follow ups (#4561)
Browse files Browse the repository at this point in the history
* model registry guide follow ups

* update images used, add a note about metrics and params on mdp

---------

Co-authored-by: Alexander Guschin <[email protected]>
  • Loading branch information
Dave Berenbaum and aguschin authored May 23, 2023
1 parent 7599e9e commit a47b15f
Show file tree
Hide file tree
Showing 9 changed files with 34 additions and 27 deletions.
3 changes: 2 additions & 1 deletion content/docs/studio/user-guide/model-registry/add-a-model.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,8 @@ your model registry, Iterative Studio creates an annotation for it in a
following ways:

1. Log your model during the training process using [dvclive].
2. Edit `dvc.yaml` directly.
2. Edit [`dvc.yaml`](/doc/user-guide/project-structure/dvcyaml-files#artifacts)
directly.
3. Use the Iterative Studio interface (watch this tutorial video or read on
below).

Expand Down
21 changes: 12 additions & 9 deletions content/docs/studio/user-guide/model-registry/assign-stage.md
Original file line number Diff line number Diff line change
Expand Up @@ -11,8 +11,8 @@ You can write CI/CD actions that can actually deploy the models to the different
deployment environments upon the creation of a new Git tag for stage assignment.
For that, you can leverage any ML model deployment tool, such as MLEM.

You can assign stages using the [`gto` CLI]. To assign stages using Iterative
Studio, watch this tutorial video or read on below:
To assign stages using Iterative Studio, watch this tutorial video or read on
below:

https://www.youtube.com/watch?v=Vrp1O5lkWBo

Expand Down Expand Up @@ -43,13 +43,16 @@ https://www.youtube.com/watch?v=Vrp1O5lkWBo

4. Optionally, provide a Git tag message.
5. Click on `Assign stage`.
6. Once the action is successful, the stage assignment will show up in the
`Stages` column of the models dashboard.
7. If you open the model details page, the stage assignment will be visible in
the model `History` section as well as in the `Stages` section.
8. If you go to your Git repository, you will see that a new Git tag referencing
the selected version and stage has been created, indicating the stage
assignment.

Once the action is successful, the stage assignment will show up in the `Stages`
column of the models dashboard.

If you open the model details page, the stage assignment will be visible in the
model `History` section as well as in the `Stages` section.

If you go to your Git repository, you will see that a new Git tag referencing
the selected version and stage has been created, indicating the stage
assignment.

[gto]: https://mlem.ai/doc/gto
[git tag]: https://git-scm.com/docs/git-tag
Expand Down
28 changes: 15 additions & 13 deletions content/docs/studio/user-guide/model-registry/register-version.md
Original file line number Diff line number Diff line change
Expand Up @@ -7,8 +7,8 @@ To register a new version of a model, Iterative Studio uses [GTO] to create an
annotated [Git tag][git tag] with the specified version number. Refer to the
[GTO docs][gto-format] to see the exact format of the Git tag.

You can register versions using the [`gto` CLI]. To register versions using
Iterative Studio, watch this tutorial video or read on below:
To register versions using Iterative Studio, watch this tutorial video or read
on below:

https://www.youtube.com/watch?v=eA70puzOp1o

Expand All @@ -31,17 +31,19 @@ https://www.youtube.com/watch?v=eA70puzOp1o

4. Optionally, provide a Git tag message.
5. Click on `Register version`.
6. Once the action is successful, the newly registered version will show up in
the `Latest version` column of the models dashboard. Note that this will
happen only if the newly registered version is the greatest semantic version
for your model. For example, if your model already had v3.0.0 registered,
then if you register a smaller version (e.g., v2.0.0), then the new version
will not appear in the `Latest version` column.
7. If you open the model details page, the newly registered version will be
available in the model `History` section as well as in the versions drop
down.
8. If you go to your Git repository, you will see that a new Git tag referencing
the selected commit has been created, representing the new version.

Once the action is successful, the newly registered version will show up in the
`Latest version` column of the models dashboard. Note that this will happen only
if the newly registered version is the greatest semantic version for your model.
For example, if your model already had v3.0.0 registered, then if you register a
smaller version (e.g., v2.0.0), then the new version will not appear in the
`Latest version` column.

If you open the model details page, the newly registered version will be
available in the model `History` section as well as in the versions drop down.

If you go to your Git repository, you will see that a new Git tag referencing
the selected commit has been created, representing the new version.

[gto]: https://mlem.ai/doc/gto
[git tag]: https://git-scm.com/docs/git-tag
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -12,7 +12,7 @@ versions using `dvc get`.
### Looking up right command in Studio

Model details page have a command ready to copy and use under "Get the model
file": ![](https://hackmd.io/_uploads/SkE5qUGr2.png)
file": ![](/img/mr-dvc-get.png)

### Constructing the command manually

Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,7 @@ of those projects that connect to this Git repository.
The models in your model registry are organized in a central dashboard that
facilitates search and discovery.

![](https://static.iterative.ai/img/studio/models-dashboard.png)
![](/img/mr-models-table.png)

You can sort the models in the dashboard by several criteria, including model
framework, repository, etc.
Expand All @@ -34,7 +34,7 @@ model registry.
You can open the details of any model in the registry by clicking on the name of
the model in the models dashboard.

![](https://static.iterative.ai/img/studio/model-details-page.png)
![](/img/mr-mdp.png)

A model details page is divided into the following sections:

Expand All @@ -57,6 +57,7 @@ A model details page is divided into the following sections:
- all assigned stages,
- version description and labels,
- path to the model, and
- metrics and params,
- model requirements and methods. Note that the requirements and methods are
identified by Iterative’s model deployment tool [MLEM]. If you have not used
MLEM, then Iterative Studio will not display any methods for the model.
Expand All @@ -81,7 +82,7 @@ belongs.
To compare model versions, select relevant commits in the project's experiment
table and click "Compare" and/or "Plots" buttons:

![](https://hackmd.io/_uploads/B1L1d8fS3.png)
![](/img/mr-comparing-models.png)

This way you can compare both registered model versions and unregistered
experimental iterations and make a decision to register a new version out of the
Expand Down
Binary file added static/img/mr-comparing-models.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added static/img/mr-dvc-get.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added static/img/mr-mdp.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file added static/img/mr-models-table.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.

0 comments on commit a47b15f

Please sign in to comment.