How Do I Renew Resources?

If the resources you purchased with the yearly/monthly billing mode have expired, renew them on the order management page. The procedure is as follows:

1. Log in to ModelArts Studio and click Order Management in the upper right corner of the page.
2. On the Order Management page, click the Resource ordering tab.
3. On the Resource ordering tab page, renew your subscriptions for data, training, and inference resources.

The following describes the pay-per-use billing rules for training jobs on ModelArts Studio.

Rules for calculating the training duration on a pay-per-use basis:

The training duration is calculated from the actual start of the training job, excluding initialization, queueing, and wait times in ModelArts Studio. The duration ends when the training job fails or is stopped.

start_time: indicates the time when the training job starts running.

duration: indicates the running duration of a training job.

Rules for special scenarios:

Retry: When a subtask of the workflow is created, the actual training duration of the new job and the actual running duration of the original job are calculated. The calculation rules are the same as the preceding rules.

Resumable training: Resumable training or retraining at a specified checkpoint is essentially a new training job. The actual running duration of the job is calculated according to the preceding rules.

Fault recovery: If fault recovery is configured and triggered in a job, and pod-level job rescheduling occurs, the duration from the start of the original job to the failure caused by the fault is recorded. If resources are sufficient during pod-level rescheduling, the new job is not queued, and the actual running duration of the new job is calculated. If resources are insufficient, the job is queued until the job status becomes Running or Failed. If queue scheduling is successful, only the actual running duration of the job is calculated.