Skip to content

Commit 1ad6ef9

Browse files
authored
Merge pull request #272207 from sdgilley/sdg-feedback
data labeling troubleshooting - access problems for v1 tabular data conversion
2 parents 36c28fe + 6b4ca64 commit 1ad6ef9

File tree

3 files changed

+39
-6
lines changed

3 files changed

+39
-6
lines changed

articles/machine-learning/how-to-manage-labeling-projects.md

Lines changed: 4 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -77,7 +77,10 @@ Additionally, when ML-assisted labeling is enabled, you can scroll down to see t
7777

7878
### Review data and labels
7979

80-
On the **Data** tab, preview the dataset and review labeled data.
80+
On the **Data** tab, preview the dataset and review labeled data.
81+
82+
> [!TIP]
83+
> Before you review, coordinate with any other possible reviewers. Otherwise, you might both be trying to approve the same label at the same time, which will keep one of you from updating it.
8184
8285
Scroll through the labeled data to see the labels. If you see data that's incorrectly labeled, select it and choose **Reject** to remove the labels and return the data to the unlabeled queue.
8386

articles/machine-learning/how-to-troubleshoot-data-labeling.md

Lines changed: 35 additions & 5 deletions
Original file line numberDiff line numberDiff line change
@@ -17,7 +17,7 @@ ms.topic: troubleshooting
1717

1818
If you have errors that occur while creating a data labeling project try the following troubleshooting steps.
1919

20-
## Add Storage Blob Data Contributor access to the workspace identity
20+
## <a name="add-blob-access"></a> Add Storage Blob Data Contributor access
2121

2222
In many cases, an error creating the project could be due to access issues. To resolve access problems, add the Storage Blob Data Contributor role to the workspace identity with these steps:
2323

@@ -48,7 +48,7 @@ In many cases, an error creating the project could be due to access issues. To r
4848
1. Select members.
4949

5050
1. In the Members page, select **+Select members**.
51-
1. Search for your workspace identity.
51+
1. Search for your workspace identity.
5252
1. By default, the workspace identity is the same as the workspace name.
5353
1. If the workspace was created with user assigned identity, search for the user identity name.
5454
1. Select the **Enterprise application** with the workspace identity name.
@@ -57,16 +57,16 @@ In many cases, an error creating the project could be due to access issues. To r
5757
:::image type="content" source="media/how-to-troubleshoot-data-labeling/select-members.png" alt-text="Screenshot shows selecting members.":::
5858

5959
1. Review and assign the role.
60-
60+
6161
1. Select **Review + assign** to review the entry.
6262
1. Select **Review + assign** again and wait for the assignment to complete.
6363

6464
## Set access for external datastore
6565

66-
If the data for your labeling project is accessed from an external datastore, set access for that datastore as well as the default datastore.
66+
If the data for your labeling project is accessed from an external datastore, set access for that datastore and the default datastore.
6767

6868
1. Navigate to the external datastore in the [Azure portal](https://portal.azure.com).
69-
1. Follow steps above starting with [Add role assignment](#add) to add the Storage Blob Data Contributor role to the workspace identity.
69+
1. Follow the previous steps, starting with [Add role assignment](#add) to add the Storage Blob Data Contributor role to the workspace identity.
7070

7171
## Set datastore to use workspace managed identity
7272

@@ -78,6 +78,36 @@ When your workspace is secured with a virtual network, use these steps to set th
7878
1. On the top toolbar, select **Update authentication**.
7979
1. Toggle on the entry for "Use workspace managed identity for data preview and profiling in Azure Machine Learning studio."
8080

81+
## When data preprocessing fails
82+
83+
Another possible issue with creating a data labeling project is when data preprocessing fails. You'll see an error that looks like this:
84+
85+
:::image type="content" source="media/how-to-troubleshoot-data-labeling/data-error.png" alt-text="Screenshot shows a data preprocessing error.":::
86+
87+
This error can occur when you use a v1 tabular dataset as your data source. The project first converts this data. Data access errors can cause this conversion to fail. To resolve this issue, check the way your datastore saves credentials for data access.
88+
89+
1. In the left menu of your workspace, select **Data**.
90+
1. On the top tab, select **Datastores**.
91+
1. Select the datastore where your v1 tabular data is stored.
92+
1. On the top toolbar, select **Update authentication**.
93+
1. If the toggle for **Save credentials with the datastore for data access** is **On**, verify that the Authentication type and values are correct.
94+
1. If the toggle for **Save credentials with the datastore for data access** is **Off**, follow the rest of these steps to ensure that the compute cluster can access the data.
95+
96+
When the **Save credentials with the datastore for data access** is **Off**, the compute cluster that runs the conversion job needs access to the datastore. To ensure that the compute cluster can access the data, find the compute cluster name and assign a managed identity, follow these steps:
97+
98+
1. In the left menu, select **Jobs**.
99+
1. Select experiment which includes the name **Labeling ConvertTabularDataset**.
100+
1. If you see a failed job, select the job. (If you see a successful job, the conversion was successful.)
101+
1. In the Overview section, at the bottom of the page is the **Compute** section. Select the **Target** compute cluster.
102+
1. On the details page for the compute cluster, at the bottom of the page is the **Managed identity** section. If the compute cluster doesn't have an identity, select the **Edit** tool to assign a system-assigned or managed identity.
103+
104+
Once you have the compute cluster name with a managed identity, assign the Storage Blob Data Contributor role to the compute cluster.
105+
106+
Follow the previous steps to [Add Storage Blob Data Contributor access](#add-blob-access). But this time, you'll be selecting the compute resource in the **Select members** section, so that the compute cluster has access to the datastore.
107+
108+
* If you're using a system-assigned identity, search for the compute name by using the workspace name, followed by `/computes/` followed by the compute name. For example, if the workspace name is `myworkspace` and the compute name is `mycompute`, search for `myworkspace/computes/mycompute` to select the member.
109+
* If you're using a user-assigned identity, search for the user-assigned identity name.
110+
81111
## Related resources
82112

83113
For information on how to troubleshoot project management issues, see [Troubleshoot project management issues](how-to-manage-labeling-projects.md#troubleshoot-issues).
9.35 KB
Loading

0 commit comments

Comments
 (0)