Resource Usage

Monitor your organization's Data Rows, Compute Credits, and Collaborator consumption in Datature Vi.

To check your organization's resource consumption in Datature Vi, open Settings, then select the Resources tab. You'll see current usage, total quotas, and renewal dates for Data Rows, Compute Credits, and Collaborators.

Before You Start
  • You need Owner, Admin, or Member access to view resource usage
  • To upgrade your plan or purchase add-ons, you need Owner access

What are resources?

Datature Vi tracks three resource types. Each one acts as a budget for a specific type of activity:

What are resources?

Resource
What it tracks
Data Rows
Asset uploads and annotation pairs
Compute Credits
GPU time for training runs and deployments
Collaborators
Team member seats

Data Rows

Data Rows measure asset storage and annotation activity. Every image you upload and every annotation pair you create draws from your Data Row quota.

What consumes Data Rows?

What consumes Data Rows?

Activity
Data Rows consumed
Upload an image
5
Create a phrase grounding pair (text + bounding box)
1
Create a VQA pair (question + answer)
1
Upload a video frame
5

Example calculation

For a phrase grounding dataset of 100 images with 50 annotation pairs:

Uploading 100 images consumes 500 Data Rows (100 x 5). Creating 50 annotation pairs consumes 50 Data Rows (50 x 1). The total for this dataset is 550 Data Rows.

Tips to reduce Data Row consumption

  • Remove duplicate or low-quality images before uploading
  • Organize your local dataset first to avoid re-uploading
  • Import pre-labeled annotations in bulk rather than creating them one at a time
  • Delete unused assets to reclaim your Data Row quota

Compute Credits

Compute Credits measure GPU time. One Compute Credit equals one minute of GPU time on a single NVIDIA T4. Higher-tier GPUs and multi-GPU configurations consume more Compute Credits per real-time minute.

GPU multipliers

Not all GPUs consume Compute Credits at the same rate. Higher-performance GPUs have higher multipliers because they provide more compute power (more VRAM, faster processing, larger model support). The T4 is the baseline at 1.0. Every other GPU is expressed as a multiple of that baseline.

Multi-GPU multipliers scale linearly with GPU count. For 4× A10G, the multiplier is 10.0 (4 × 2.5).

Full GPU multiplier table

GPU type
VRAM
CUDA cores
Multiplier
Available counts
NVIDIA T4
16 GB
2,560
1.0×
1, 4, 8 GPUs
NVIDIA L4
24 GB
7,680
2.0×
1, 4, 8 GPUs
NVIDIA A10G
24 GB
9,216
2.5×
1, 4, 8 GPUs
NVIDIA A100 (40 GB)
40 GB
6,912
4.0×
8 GPUs
NVIDIA A100 (80 GB)
80 GB
6,912
6.0×
8 GPUs
NVIDIA H100
80 GB
14,592
12.0×
1, 8 GPUs

Example calculation

A training run on 4× A10G GPUs for 45 minutes:

The 4× A10G multiplier is 10.0. Over 45 minutes, that totals 450 Compute Credits (10.0 × 45).

What consumes Compute Credits?

What consumes Compute Credits?

Activity
Consumption
Model training run
Per minute of GPU time
Model deployment
Per minute of inference serving

Estimating Compute Credit costs

30-minute run estimates

GPU configuration
Compute Credits
1× T4
30
4× T4
120
1× A10G
75
8× A100 (40 GB)
960
1× H100
360

2-hour run estimates

GPU configuration
Compute Credits
1× T4
120
4× T4
480
1× A10G
300
8× A100 (40 GB)
3,840
1× H100
1,440

8-hour run estimates

GPU configuration
Compute Credits
1× T4
480
4× T4
1,920
1× A10G
1,200
8× A100 (40 GB)
15,360
1× H100
5,760

Tips to reduce Compute Credit consumption

  • Start with a single T4 to validate your training configuration before scaling up
  • Enable early stopping to end runs that are no longer improving
  • Cancel runs that show no sign of converging rather than waiting them out
  • Use T4 or L4 for experimentation and A100 or H100 for production runs

Collaborators

Collaborators track team member seats. Each person you invite uses one slot. Removing a member frees that slot immediately.

What affects Collaborator count?

Activity
Change
Invite a new member
−1 available slot
Remove a member
+1 available slot

Example: If your plan allows 10 Collaborators and you currently have 7 members, you have 3 available slots. Inviting 2 new members brings usage to 9 slots with 1 remaining. Removing 1 member afterward brings it to 8 slots used with 2 available.

To manage team members, see Add members and Remove members.


Reading the Resources page

The Resources page shows three indicators for each resource type:

  • Current consumption: How much you've used
  • Total quota: Your plan's limit
  • Progress bar: Visual percentage of usage used

At the bottom right is your plan's monthly renewal date. This is when your Compute Credits reset. Data Rows and Collaborators are fixed allocations, so they will not reset each month.

Usage updates in real time as you upload assets, create annotations, start runs, deploy models, and invite or remove members.


Plan quotas

Resource quotas depend on your subscription plan:

Plan
Data Rows
Compute Credits
Collaborators
Free
3,000
300
Solo use only
Developer
10,000
Pay-per-use
10
Professional
50,000
5,000
50
Enterprise
Custom
Custom
Custom

The table above shows example allocations. Your actual quotas depend on your specific subscription. Check the Resources page for your current plan and quotas.

Renewal schedule

  • Data Rows: fixed to your plan; do not renew monthly
  • Compute Credits: renew monthly on your plan anniversary date
  • Collaborators: fixed slots; do not renew monthly

What happens when you hit a limit?

What happens when you hit a limit?

Resource
Effect
Resolution
Data Rows
Cannot upload assets or create annotation pairs
Upgrade plan or purchase add-ons
Compute Credits
Cannot start training runs or deployments
Wait for monthly renewal or purchase add-ons
Collaborators
Cannot invite new members
Remove inactive members or upgrade plan

Common troubleshooting

Check your Data Row balance on the Resources page. If you're at quota, upgrade your plan, purchase a Data Row add-on, or delete unused assets and annotation pairs so those Data Rows return to your balance.

Check your Compute Credit balance. If Compute Credits are depleted, you can:

  1. Cancel any active runs that aren't needed
  2. Wait for your monthly renewal
  3. Purchase a Compute Credit add-on
  4. Reduce the GPU count or switch to a lower-tier GPU

You've likely reached your Collaborator limit. Options:

  1. Remove inactive or former members to free slots
  2. Revoke pending invitations that haven't been accepted
  3. Upgrade to a plan with more Collaborator slots

Frequently asked questions

Data Rows and Collaborators are fixed to your plan and don't roll over. Compute Credits typically reset monthly; unused Compute Credits don't carry forward. Check your plan terms for specifics.

Some plans offer add-on packs for Data Rows and Compute Credits. Contact [email protected] to check what's available for your plan.

Your existing data stays safe. You can still view and work with what you have. New operations that require the depleted resource will fail until you renew or upgrade.

Yes. When you delete assets or remove annotation pairs from your dataset, the corresponding Data Rows are immediately returned to your available balance and can be reused. Review your Data Row usage after major deletions to confirm your updated quota.

No. Resources are scoped to a single organization and cannot be transferred or shared between organizations.


Upgrading your plan

Consider upgrading when:

  • You're consistently reaching 80%+ of any quota
  • You're planning a larger project that will exceed current allocations
  • Your team needs more Collaborator slots
  • You're training larger models or running more frequent experiments

To upgrade, click Change Plan on the Resources page, or contact [email protected] for Enterprise plans and custom quotas.


Next steps

Add Team Members

Invite Collaborators and configure their access to datasets and training projects.

Secret Keys

Create API credentials to authenticate SDK and programmatic access.

Team Settings

Manage organization members, roles, and access controls.