Skip to main content
torch.jstorch.jstorch.js
Getting StartedPlaygroundContact
Login
torch.jstorch.jstorch.js
Documentation
IntroductionType SafetyTensor IndexingEinsumEinopsAutogradTraining a ModelProfiling & MemoryPyTorch MigrationBest PracticesRuntimesPerformance
IntroductionQuickstartClient APIWorker APIFile I/OBackends
torch.js· 2026
LegalTerms of UsePrivacy Policy
  1. docs
  2. Spark
  3. Client API

Client API

The Spark Client API provides the React hooks and controls necessary to manage your background training workers from your main user interface.

spark.use()

The primary entry point for React applications. It manages the worker lifecycle and provides reactive bindings.

function use<T>(workerFn: () => void): SparkProxy<T>;

Example Usage

const s = spark.use(myWorker);

// Access reactive state
const loss = s.state.loss.value;

// Call worker functions
const onStart = () => s.train(100);

The Controller (s.ctrl)

The ctrl object gives you direct control over the execution state of the current worker function.

MethodAction
pause()Suspends execution at the next spark.checkpoint()
resume()Resumes a previously paused worker
stop()Terminates the current function execution immediately
<button onClick={() => s.ctrl.stop()}>Emergency Stop</button>

Managing Loading States

Because workers are initialized asynchronously, the s.state object might be undefined for the first few frames. Always use optional chaining or provide default values.

const accuracy = s.state?.accuracy?.value ?? 0;

Next Steps

  • Worker API - See how to expose values from the other side of the bridge.
  • File I/O - Connect your client to the cloud.
Previous
Quickstart
Next
Worker API