can1357/oh-my-pi · error · Error

No stt model devices configured

Error message

No stt model devices configured

What it means

When loading a transformers-engine STT model, loadPipelineWithDeviceFallback iterates tinyModelDeviceLoadOrder (the user's preferred device plus fallbacks such as webgpu/metal/cpu) and tries each; failures are logged per device. If every device in the load order fails, or the order itself is empty, it throws 'No stt model devices configured'.

Source

Thrown at packages/coding-agent/src/stt/asr-worker.ts:219

		const device = devices[i]!;
		try {
			return {
				pipeline: await loadPipelineOnDevice(transformers, spec, modelKey, transport, requestId, device),
				device,
			};
		} catch (error) {
			if (i === devices.length - 1) throw error;
			const fallbackDevice = devices[i + 1]!;
			sendLog(transport, "warn", "stt: accelerated device failed; falling back", {
				modelKey,
				repo: spec.repo,
				device,
				fallbackDevice,
				error: errorMessage(error),
			});
		}
	}
	throw new Error("No stt model devices configured");
}

async function loadTransformersModel(
	spec: TransformersSttModelSpec,
	modelKey: SttModelKey,
	transport: SttTransport,
	requestId: string,
): Promise<LoadedModel> {
	const transformers = await loadTransformersRuntime(
		transformersRuntime,
		transport,
		requestId,
		modelKey,
		getSttRuntimeDir,
	);
	const startedAt = performance.now();
	const { pipeline, device } = await loadPipelineWithDeviceFallback(
		transformers,

View on GitHub (pinned to 9690622007)

Solutions

1. Check the per-device warn/error logs emitted before this throw to see why each device failed.
2. Update GPU drivers / ensure WebGPU (browser-grade) or Metal support, or force device=cpu via the tiny-model device preference config.
3. Clear the transformers model cache and retry in case a corrupted download caused each device load to fail.
4. Update @huggingface/transformers and onnxruntime to compatible versions.

const devices = tinyModelDeviceLoadOrder(preference);
if (devices.length === 0) throw new Error("no valid stt devices; check tiny-model device config");

try { const { pipeline, device } = await loadPipelineWithDeviceFallback(...); } catch (err) { // inspect logged per-device errors; surface guidance to fix drivers or force CPU }

When it happens

Trigger: Loading a transformers Whisper-tier model when every candidate device (GPU and CPU) fails to create the ASR pipeline, or when tinyModelDeviceLoadOrder returns an empty list for the resolved device preference.

Common situations: GPU runtime unavailable (missing WebGPU/Metal drivers) combined with a broken CPU fallback, an exotic device preference in config that yields no valid devices, or transformers.js unable to load the ONNX backend on any device.

Related errors

• No such device or address
• Unable to resolve ${SHERPA_PACKAGE} in compiled runtime at $
• Unknown stt model: ${modelKey}
• Failed to download speech model (${spec.repo})${detail}
• Speech model download finished without required files (${spe

AI-assisted analysis of can1357/oh-my-pi@9690622007 (2026-08-31). Data as JSON: /api/errors/39f2162295b1f071. Report an issue: GitHub.