Quota Management
How to ensure resources are reasonably allocated and used when platform resources are limited?
Typical scenarios:
- GPU resources are limited, need to control the number of training tasks
- Storage and bandwidth are limited, need to limit data export frequency
- AI resources are limited, need to control automatic annotation feature usage
- Different user levels need different resource quotas
Quota management is designed to solve these problems. By configuring independent quota limits for each user, it can control resource usage, prevent resource abuse, and ensure stable system operation.
Quota Types
The platform supports four types of quota management:
Training Quota
Controls the number of times or duration users can perform model training.
Use cases:
- Limit the number of training tasks to prevent excessive GPU resource consumption
- Control GPU usage time to optimize resource allocation
- Allocate training resources by user level
Quota Display:
- Total Quota: Total number of training times or duration available to users
- Used: Currently used training times or duration
- Remaining Quota: Total quota minus used quota
Inference Quota
Controls the number of inference services users can create and run.
Use cases:
- Limit the number of concurrent inference services
- Control inference resource usage
- Allocate inference resources by project
Quota Display:
- Total Quota: Total number of inference services users can create
- Used: Currently created inference service count
- Remaining Quota: Total quota minus used quota
Export Quota
Controls the number of times or data volume users can export data.
Use cases:
- Limit data export frequency to prevent excessive exports
- Control storage and bandwidth resource usage
- Allocate export resources by user level
Quota Display:
- Total Quota: Total number of export times or data volume available to users
- Used: Currently used export times or data volume
- Remaining Quota: Total quota minus used quota
Auto Annotation Quota (New in 3.3.0)
Controls the number of times users can use the automatic annotation feature.
Use cases:
- Limit automatic annotation feature usage to control AI resource consumption
- Allocate automatic annotation resources by user level
- Prevent abuse of automatic annotation feature
Quota Display:
- Total Quota: Total number of automatic annotation times available to users
- Used: Currently used automatic annotation count
- Remaining Quota: Total quota minus used quota
Quick Start: Configure Quota for Users
Step 1: Select Quota Type
Go to the Quota Management page and select the quota type to configure:
- Training Quota
- Inference Quota
- Export Quota
- Auto Annotation Quota
Step 2: Create Quota Configuration
- Click "New" button
- Select User: Choose the user to configure quota from the user list
- Set Total Quota: Enter the user's total quota amount
- Add Note (Optional): Explain the reason or purpose of quota configuration
- Save configuration
Step 3: View Quota Usage
In the quota list you can see:
- User name and role
- Total quota and used quota
- Remaining quota
- Quota status (Normal/Insufficient/Exhausted)
Advanced Usage
How to Set Global Quota?
Global Configuration Features:
- Applies to all users
- Serves as default quota configuration
- User-specific configuration takes priority over global configuration
Setting Method:
- Set global quota on the Quota Management page
- Global configuration will be displayed in user quota information
- Can set independent configuration for specific users to override global configuration
Use Cases:
- Set basic quota for all users
- Set higher quota for specific users
- Simplify quota management and reduce duplicate configuration
How to Adjust User Quota?
Adjustment Method:
- Find the user to adjust in the quota list
- Click "Edit" button
- Modify total quota amount
- Save changes
Notes:
- Modifying total quota will not affect used quota
- If new quota is less than used quota, users will not be able to continue using that feature
- Recommend notifying users before modification
How to View Quota Usage Statistics?
Statistics Information:
- Total user count and configured user count
- Total quota and total usage
- Average usage rate
- Number of users with insufficient quota
Query Functions:
- Search by user
- Filter by quota type
- View quota usage trends
- Export quota configuration list
Quota Control Mechanism
How Are Restrictions Applied?
When user quota is insufficient, the system will automatically block related operations:
Restriction Behavior:
- Training Tasks: Cannot create new training tasks
- Inference Services: Cannot create new inference services
- Data Export: Cannot perform export operations
- Auto Annotation: Cannot use automatic annotation feature
Prompt Information:
- Clearly display the reason for insufficient quota
- Display current quota and remaining quota
- Prompt to contact administrator to increase quota
Quota Consumption Rules
Each time related features are used, the system automatically deducts corresponding quota:
Consumption Rules:
- Training Tasks: Deducted when created, regardless of whether task succeeds
- Inference Services: Deducted when created, quota not returned when service is deleted
- Data Export: Deducted when export succeeds
- Auto Annotation: Deducted when usage succeeds
Quota Return:
- Generally, used quota will not be returned
- Special cases can contact administrator for manual adjustment
Use Cases
Scenario 1: Allocate Resources by User Level
Configuration Steps:
- Set different quotas for users of different levels
- Set higher quota limits for VIP users
- Set basic quota limits for regular users
- Set trial quota for new users
Example:
- VIP Users: Training quota 100 times, Inference quota 20
- Regular Users: Training quota 10 times, Inference quota 2
- Trial Users: Training quota 3 times, Inference quota 1
Scenario 2: Allocate Resources by Project
Configuration Steps:
- Set overall quota limit for projects
- Project members share project quota
- Monitor project resource usage
- Adjust quota based on project progress
Scenario 3: Resource Usage Monitoring
Monitoring Method:
- Regularly view user quota usage
- Identify users with abnormal resource usage
- Optimize resource allocation strategies
- Prevent resource abuse
Common Questions
What to Do When Quota is Insufficient?
Solution:
- View current quota and remaining quota
- Contact administrator to apply for quota increase
- Wait for administrator to adjust quota before continuing to use
How to View Your Own Quota?
Viewing Method:
- Training Page: Display training quota usage
- Inference Page: Display inference quota usage
- Export Page: Display export quota usage
- Auto Annotation Feature: Display auto annotation quota usage
Can Quota Be Restored?
Restoration Note:
- Generally, used quota will not be returned
- Special cases can contact administrator for manual adjustment
- Deleting training tasks or inference services will not return quota
How to Apply for Quota Increase?
Application Method:
- View current quota usage
- Contact platform administrator
- Explain application reason and required quota
- Wait for administrator review and adjustment
Applicable Roles
Administrator
You can:
- Configure and manage quota for all users
- Set global quota configuration
- Monitor overall resource usage
- Adjust user quota based on needs
- View quota usage statistics and reports
User
You can:
- View your own quota usage
- Understand quota limits for each feature
- Reasonably plan resource usage
- Contact administrator when quota is insufficient
Notes
- Quota Cannot Be Restored: Used quota generally will not be returned, please use carefully
- Insufficient Quota Handling: When quota is insufficient, corresponding features cannot be used, need to contact administrator to increase quota
- Global Configuration Priority: User-specific configuration takes priority over global configuration
- Quota Statistics: Quota statistics may have delays, actual usage is authoritative
Related Features
After completing quota management, you may also need:
- User Management: Manage user accounts
- Model Training: View training quota usage
- Inference Service: View inference quota usage