yum-archive/TaSTT-Whisper
High-performance GPGPU inference of OpenAI's Whisper automatic speech recognition (ASR) model
git clone https://git.yummers.dev/yum-archive/TaSTT-Whisper
11c399b
master
1namespace Whisper 2{ 3/// <summary>Implementation value for the <see cref="Library.loadModel(string, eGpuModelFlags, eModelImplementation)" /> factory function</summary> 4public enum eModelImplementation : uint 5{ 6/// <summary>GPGPU implementation based on Direct3D 11.0 compute shaders</summary> 7GPU = 1 , 8 9/// <summary>A hybrid implementation which uses DirectCompute for encode, and decodes on CPU</summary> 10/// <remarks> 11/// <para>The build of the native DLL included into this nuget package doesn’t implement this version.<br/> 12/// To enable, edit <c>stdafx.h</c> in Whisper project, change the value of <c>BUILD_HYBRID_VERSION</c> macro from zero to one, and build.</para> 13/// <para>This implementation requires a CPU with AVX1, FMA3, F16C and BMI1 instruction set extensions.</para> 14/// </remarks> 15Hybrid = 2 , 16 17/// <summary>A reference implementation which uses the original GGML CPU-running code.</summary> 18/// <remarks> 19/// <para>The build of the native DLL included into this nuget package doesn’t implement this version either.<br/> 20/// To enable, edit <c>stdafx.h</c> in Whisper project, change the value of <c>BUILD_BOTH_VERSIONS</c> macro from zero to one, and build the project.</para> 21/// <para>This implementation requires a CPU with AVX1, FMA3, and F16C instruction set extensions.</para> 22/// </remarks> 23Reference = 3 , 24} 25}