Inference & Serving · advanced
Browser ML Runtime
Web Workers, WASM, OPFS, TypedArrays.
Mental model
You can run ML in the browser using libraries like ONNX Runtime Web, TensorFlow.js, or Transformers.js. You trade server cost and user privacy for a slower first load and a limit on model size. WebGPU is fastest, WASM works everywhere, CPU is the fallback. Scope: this card owns the browser as a platform — Workers, WASM, OPFS, model loading and the first-load budget. The GPU backend specifically is `ml-webgpu`; native and mobile execution is `local-on-device-inference`.
How to study Browser ML Runtime
Begin by restating the mental model in your own words, then connect it to a concrete system you have built or operated. Name the mechanism, the constraint it addresses, and the trade-off it introduces. Use Transformers.js (Hugging Face) to check details, but close the source before writing your explanation. Retrieval is the learning step; rereading is only preparation.
Next, compare Browser ML Runtime with WebGPU Compute. Ask what changes in correctness, latency, resource use, operability, and failure recovery. Complete WASM linear memory growth and preserve the command, input, output, and one failed attempt as evidence. Finish by explaining the idea without jargon to someone who has not studied the track.
Proof of understanding
- Explain the mechanism from first principles and identify the state it reads or changes.
- Give one situation where the concept is the right choice and one where it is not.
- Predict a realistic failure mode before running the drill, then compare the prediction with evidence.
- Connect the result to a roadmap or build artifact instead of treating the concept as isolated trivia.
Learn from primary sources
Practice and explain it back
WASM linear memory growth
Page limit ~2–4GB WASM memory. Model weights 1.5B params float16 ≈ 3GB. Can you load weights + activations in one page tab without sharding?
Expected evidence: No headroom — need quantization, offloading, or worker sharding.
Open the interactive drill →Review prompts
- Why does the main thread have to be kept out of browser inference, and what does that force into the design?
Build evidence
Synthesize: Inference & Serving
Build a production mental model for inference engines, memory, kernels, routing, hardware, and serving economics. Produce one working system, benchmark, or evidence-backed design that integrates the path.
- Implements or precisely models the core mechanisms from all three milestones
- Includes at least one injected failure or adversarial case and demonstrates recovery
- Reports quality, latency, resource, reliability, or usability measurements relevant to the domain
- Ships a concise architecture note explaining decisions, trade-offs, and remaining risks