aws / aws/sagemaker-python-sdk

TensorBoard get_url feature not creating proper redirect link to SageMaker Studio

Offen
#5,252 0 Kommentare 0 Reaktionen 0 zugewiesene Personen Auf GitHub ansehen
component: pysdk-team type: bug
Vorherrschende Sprache
Python
Sterne
2.3k
Forks
1.3k
Ø Merge
1 T. 22 Std.
Gemergte PRs (30 T.)
35

Beschreibung

# TensorBoard get_url feature not creating proper redirect link to SageMaker Studio

## Describe the bug
The `get_app_url` method in the TensorBoard application class (`sagemaker.interactive_apps.tensorboard.TensorBoardApp`) does not create proper redirect links when used with `create_presigned_domain_url=True`. Currently, the generated URL redirects to a SageMaker Studio blank page instead of opening the TensorBoard application directly. This makes it difficult to access TensorBoard visualizations from outside the Studio environment.

## To reproduce

The issue can be reproduced with the following code (extract from the [documentation](https://docs.aws.amazon.com/sagemaker/latest/dg/debugger-htb-access-tb-url.html#debugger-htb-access-tb-url-presigned)):

```python
from sagemaker.interactive_apps import tensorboard
region = "us-east-1"
app = tensorboard.TensorBoardApp(region)

print(
app.get_app_url(
create_presigned_domain_url=True, # Required to be set to True for creating a presigned URL
domain_id="your-domain-id", # Replace with your actual domain ID
user_profile_name="your-user-profile", # Replace with your actual user profile name
open_in_default_web_browser=False, # Set to False to print the URL to terminal
)
)
```

When executing this code, the generated URL opens a blank SageMaker Studio page instead of directly opening the TensorBoard application.

## Expected behavior

The generated URL should directly open the TensorBoard application in SageMaker Studio. The `LandingUri` parameter in `optional_create_presigned_url_kwargs` should correctly route the user to the TensorBoard application with the proper URI format.

The correct behavior would be for the URL to open directly to the TensorBoard interface, bypassing the need for manual navigation within Studio.

### How to make it work

After investigating this issue, I found that the correct format for the `LandingUri` parameter needs to be specified differently to properly redirect to TensorBoard. Here's the working solution:

```python
from sagemaker.interactive_apps import tensorboard
region = "us-east-1"
app = tensorboard.TensorBoardApp(region)

print(
app.get_app_url(
create_presigned_domain_url=True, # Required to be set to True for creating a presigned URL
domain_id="your-domain-id", # Replace with your actual domain ID
user_profile_name="your-user-profile", # Replace with your actual user profile name
open_in_default_web_browser=False, # Set to False to print the URL to terminal
optional_create_presigned_url_kwargs={
"LandingUri": "app:TensorBoard:default/"
}
)
)
```

## Screenshots or logs

When accessing the generated URL, you're taken to a blank SageMaker Studio page instead of the TensorBoard application.

Image

## System information
- **SageMaker Python SDK version**: 2.249.0 - latest
- **Framework name**: TensorFlow
- **Framework version**: N/A
- **Python version**: 3.13
- **CPU or GPU**: N/A
- **Custom Docker image**: N/A

## Additional context

The issue appears to be in how the TensorBoard application integration handles the `LandingUri` parameter from the underlying `create_presigned_domain_url` API.

A potential fix would be to update the SDK to properly handle the TensorBoard-specific URI format in the `get_app_url` method of the TensorBoardApp class.

The AWS documentation needs to be update or the SDK fixed

Beitragsleitfaden

Beitragsleitfaden öffnen

Rechercherichtung

Beginne mit sagemaker.interactive_apps.tensorboard.TensorBoardApp.get_app_url und verfolge, wie optional_create_presigned_url_kwargs die LandingUri bereitstellt, wenn create_presigned_domain_url=True ist. Reproduziere das bereitgestellte Beispiel und überprüfe, dass die generierte URL die TensorBoard-Anwendung direkt in SageMaker Studio öffnet, statt eine leere Seite anzuzeigen.

Vom Indexierungsmodell aus dem Issue-Text verfasst.

Bewertung

Tech-Stack
aws, python
Bereich
cloud, machine-learning
Issue-Typ
Bug
Schwierigkeit
2/5
Geschätzter Aufwand
1-3 Stunden
Aktivitätsstatus
Veraltet
Klarheit
Klar beschrieben
Anfängerfreundlichkeit
45/100

Neue Issues direkt in Ihr Postfach

Eine kurze Übersicht über anfängerfreundliche GitHub-Issues.