phonon-2-webgpu / validation.json
alperugurcan's picture
Add ONNX-free Phonon-2 WebGPU transcription app
313d50a verified
Raw History Blame Contribute Delete
2.56 kB
{
"backend": "wgpu-native / Vulkan / llvmpipe software adapter",
"execution": "Unmodified integrated engine.js, actual WGSL compute, JavaScript frontend",
"cases": [
{
"audioSeconds": 1,
"text": "He hoped",
"tokens": 4,
"frames": 13,
"textMatch": true,
"tokenIdsMatch": true,
"tokenFramesMatch": true
},
{
"audioSeconds": 10.435,
"text": "He hoped there would be stew for dinner, turnips and carrots and bruised potatoes, and fat mutton pieces to be ladled out in thick, peppered, flour fattened sauce,",
"tokens": 56,
"frames": 131,
"textMatch": true,
"tokenIdsMatch": true,
"tokenFramesMatch": true
}
],
"limitations": [
"User confirmed browser microphone transcription; exact hardware-browser reference comparison not reported",
"Reference is independent NumPy implementation, not native Phonon",
"Timing on software adapter is not a hardware browser benchmark"
],
"repeatedTranscription": {
"first": {
"run": 1,
"text": "He hoped",
"fetches": 27,
"pipelines": 27,
"devices": 1,
"seconds": 21.127447283
},
"second": {
"run": 2,
"text": "He hoped",
"fetches": 0,
"pipelines": 0,
"devices": 1,
"seconds": 19.427448858999995
}
},
"preparedGpuWeights": {
"residentBytes": 2508229144,
"warmRun": {
"first": {
"run": 1,
"text": "He hoped",
"fetches": 27,
"pipelines": 27,
"devices": 1,
"weightWrites": 273,
"seconds": 21.49359271
},
"current": {
"run": 2,
"text": "He hoped",
"fetches": 0,
"pipelines": 0,
"devices": 1,
"weightWrites": 0,
"seconds": 1.8481516139999985
}
},
"restoreLowMemory": {
"first": {
"run": 1,
"text": "He hoped",
"fetches": 27,
"pipelines": 27,
"devices": 1,
"weightWrites": 273,
"seconds": 21.49359271
},
"current": {
"run": 3,
"text": "He hoped",
"fetches": 0,
"pipelines": 0,
"devices": 1,
"weightWrites": 273,
"seconds": 19.569735493999996
}
}
},
"browserConfirmation": {
"source": "User report in this chat",
"input": "Microphone recording",
"audioSeconds": 5.16,
"reportedProcessingSeconds": 20.54,
"text": "Hey, what's up, my friend? Yeah, yeah, yeah.",
"exactReferenceComparison": false
}
}