Resource Usage
Resource Usage page from Workspace Settings (click image to enlarge)
Checking Resource Usage
Check your account's quota and usage for the current month by following these steps:
-
Select Your Workspace - On your Nexus Dashboard, you can select the dropdown menu for workspaces and find the workspace associated with you.
-
Select Workspace Settings - This is in the top right button on the dashboard for Settings.
-
Select Resource Usage - This can be found on the left sidebar.
For each resource, you will be able to view your current usage, and your quota (the maximum amount you can use under your plan). We describe each resource type below.
| Resource Type | Description | Resetting Usage |
|---|---|---|
| Compute Minutes | Number of GPU computing minutes. This is calculated relative to a single Nvidia T4 GPU. More powerful GPUs use up more Compute Minutes. | Automatically reset to 0 monthly |
| Images Uploaded | Number of images uploaded, across all projects | Delete unneeded images to reduce usage |
| Artifacts Stored | Number of artifacts stored in the Artifacts page, across all projects | Delete unneeded artifacts to reduce usage |
| Artifacts Exported | Number of TensorFlow models downloaded from the Artifacts page, across all projects | Automatically reset to 0 monthly |
| IntelliBrush Tokens Used | Number of annotations created created using the Annotator's Intelligent Tools, including'IntelliBrush', 'IntelliMatch','Everything'features and annotations created with 'Tracking Mode' enabled. | Automatically reset to 0 monthly |
Please note that -
- Training will stop once your account hits the quota for Compute Minutes, even if model training is ongoing
- Updates to your model will no longer be stored as artifacts once your account hits the quota for Artifacts Stored, even if model training is ongoing
To increase your usage quota, consider upgrading your plan.
Compute Minutes Calculation
Compute Minutes are calculated using the following formula:
Real-time duration of training * Number of GPUs used * GPU Usage Multiplier = Compute Minutes,
where real-time duration of training is the number of minutes that the model trains for in real time, number of GPUs is the number of GPUs selected in running the workflow in Training Option : Hardware Acceleration, and the GPU Usage Multiplier is based on the type of GPU used, also selected in Training Option : Hardware Acceleration. The GPU usage multiplier for the types of GPUs offered is displayed below:
| GPU Type | GPU Specifications | GPU Usage Multiplier |
|---|---|---|
| Nvidia T4 | - 2560 CUDA Cores - 16 GB Memory | 1 |
| Nvidia P100 | - 3584 CUDA Cores - 16 GB Memory | 3 |
| Nvidia V100 | - 5120 CUDA Cores - 16 GB Memory | 4 |
| Nvidia L4 | - 7680 CUDA Cores - 24 GB Memory | 2.2 |
| Nvidia A100 | - 6912 CUDA Cores - 40 GB Memory | 6 |
Examples of Compute Minutes Calculations
The real-time training duration is multiplied with each of the multipliers to end up with the resulting compute minutes as shown below:
| Real-time training duration (minutes) | GPU Type | GPU Usage Multiplier | Number of GPUs selected | Compute Minutes |
|---|---|---|---|---|
| 30 | Nvidia T4 | 1 | 1 | 30 * 1 * 1 = 30 |
| 60 | Nvidia T4 | 1 | 4 | 60 * 1 * 4 = 240 |
| 30 | Nvidia P100 | 3 | 2 | 30 * 3 * 2 = 180 |
| 60 | Nvidia P100 | 3 | 4 | 30 * 3 * 4 = 360 |
Updated 19 days ago