3 min read
Training & Metrics
How training runs in DLWAY: in a Web Worker on your own GPU, with loss, accuracy and resource usage streaming to the Dashboard.
Live metrics
Training runs in a dedicated Web Worker. Loss, accuracy, and resource usage (CPU, memory, GPU context) stream live to the Dashboard as training progresses.
Checkpoints
Save checkpoints during a run and resume later, or compare metrics across runs directly in the Dashboard.
Where the compute comes from
In-browser training uses TensorFlow.js on your own GPU through WebGPU, falling back to WebGL. For native Python frameworks, attach a Jupyter kernel under Settings → Compute.
See Remote compute for what runs on the kernel and what the limits are.