Can DGX Spark handle YOLO training with 100 classes and 150k images (640x640, 200 epochs, batch 32

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
1

Hi NVIDIA team,

I’m evaluating the NVIDIA DGX Spark system for large-scale training.
I’d like to confirm whether it’s capable of efficiently training a YOLO (v8/11) model with the following parameters:

  • Dataset: 100 classes

  • Images: ~150,000 (640×640)

  • Batch size: 32

  • Epochs: 200

  • Framework: PyTorch (using Ultralytics or custom YOLO setup)

Could you please provide:

  1. Estimated training time on DGX Spark (with default configuration).

  2. Recommended GPU/memory configuration.

  3. Any specific settings or optimizations for this workload.

2

We haven’t tested YOLO on the Spark but it should have enough memory for your usecase