Platform For AI:FAQ about algorithm components

Which data type fields of MaxCompute are supported by the algorithm components of Machine Learning Designer?

Only BOOLEAN, BIGINT, DOUBLE, STRING, and DATETIME are supported.

How do I use the Beta algorithm components?

Machine Learning Designer provides custom Beta algorithm components, such as text approximate similar pairs and multi-hot encoder train, to meet specific business requirements. We recommend that you use other components that have similar functions or submit a ticket to obtain more information.

What do I do if the x13-auto-arima component reports an error?

• Error message

  row number should be (1, 1200].

• Solution

Make sure that each subset of the training dataset that is imported into x13-auto-arima contains no more than 1,200 entries, or this training dataset contains no more than 1,200 entries.

What do I do if the Doc2Vec component reports the CallExecutorToParseTaskFail error?

Make sure that the dataset that is imported into the Doc2Vec component contains less than 2,410,000 × 10,000 entries. This limit is calculated based on the following formula: (Number of documents + Number of words) × Vector length. Make sure that the number of users is less than 42,432,500 × 7,712,293 × 300. If the upper limits are exceeded, memory requests may fail.

You can reduce the size of the dataset and then perform word segmentation on the dataset before you import the dataset.

How do I persist temporary tables that are generated by the nodes on the canvas after I run the nodes?

Machine Learning Designer provides the Write Table component. To persist temporary tables, you must connect the Write Table component as a downstream node to the node that generates the temporary tables.

What do I do if a component reports an error that the RAM user is not authorized to access MaxCompute?

In the workspace, check whether the Resource Access Management (RAM) user that you use has the MaxCompute Developer role added. For more information about the procedure, see Manage the members of a workspace.

What do I do if the column selector of a component cannot find the specified column?

Check whether the upstream node successfully generated the data table that is required by the current node by right-clicking the upstream node and then selecting Stop Here in the shortcut menu.

What do I do if deep learning-related components report an error that no resources can be allocated?

If the following error message is reported: ODPS-0432151: Quota not enough - You have no allocable Fuxi quota by now, all your GPU resources are occupied. You can purchase or scale a GPU-accelerated subscription resource group, or wait until the resources are released from the current tasks.

What do I do if deep learning-related components report an error that no GPU resources are available?

If your workspace is deployed in the China (Beijing), China (Shanghai), China (Hangzhou), or China (Shenzhen) region and the following error message is reported: No available cluster has the required feature: fuxi_gpu,gpu_with_vm_cuda8, the GPU resources of the MaxCompute project are not enabled for the workspace. Currently, MaxCompute resources do not support GPU jobs. We recommend that you use DLC resources. For example, you can migrate jobs by using an Python Script component.

What do I do if a pipeline fails to run and reports an error that I am not in the MaxCompute project?

In the Members panel of the Workspace Details page, assign the MaxCompute Developer role to the RAM user that you use. For more information, see Manage the members of a workspace.

How do I identify the issue if an algorithm component fails to run?

1. On the canvas of the Machine Learning Designer pipeline, right-click the component that failed to run. In the shortcut menu, select View Log.

2. On the log tab, find and click the logview link.

3. On the Job Details tab, click the icon in the StdErr column.

4. In the StdErr dialog box, view the error log information.