Deployment center
Every path records fit, verified versions, startup commands, production responsibilities, and known limitations.
Run Fun-ASR-Nano on NVIDIA GPUs for file batches or an OpenAI-compatible transcription endpoint.
Throughput and memory depend on segmentation, batch size, concurrency, GPU, and vLLM version; retest with target traffic.
Use cross-platform release packages or source builds to run FunASR GGUF models on CPUs, desktop GPUs, and edge devices.
Prebuilt GPU packages cover only the documented backend and architecture; other devices require a source build and hardware validation.
Serve FunASR models through `/v1/audio/transcriptions` for existing SDKs, agents, and workflows.
The example server does not include production authentication, tenant quotas, or full observability; provide them at the gateway and infrastructure layers.
Use online or two-pass WebSocket protocols for microphones, meeting captions, and continuous speech streams.
Capacity depends on realtime ratio, chunks, VAD, connection count, and client backpressure; file RTFx does not replace concurrency tests.
Validate with Docker Compose, then deploy with the Kubernetes template, persistent model cache, probes, and private ClusterIP.
The CPU example is not a CUDA production image; GPU drivers, image, scheduling, and capacity still require platform validation.
Deploy offline, online, and two-pass ASR on general CPUs with ONNX Runtime and C++ paths.
Model conversion, operator support, and thread settings vary by platform; measure capacity on the target CPU with real audio.
Move from a runnable model to an observable, scalable, reversible production service with explicit boundaries.
A checklist does not replace workload, accuracy, failure, and security exercises for the target product.
