akiki-vad

Akiki-AI 's Collections

updated 16 days ago

Akiki VAD base model, five linked safetensors precision variants, and training-data provenance. RealFFT remains FP32.

Akiki-AI/akiki-vad-1.0

Audio Classification • 277k • Updated 16 days ago • 33

Note Original Akiki VAD 1.0 model. Five separate precision variants are linked through its Quantizations tree.
Akiki-AI/akiki-vad-1.0-fp16

Audio Classification • 190k • Updated 16 days ago • 32

Note FP16 quantization of Akiki VAD 1.0. Unchanged FP32 RealFFT; validated on TEN and VoxConverse.
Akiki-AI/akiki-vad-1.0-bf16

Audio Classification • 190k • Updated 16 days ago • 33

Note BF16 quantization of Akiki VAD 1.0. Unchanged FP32 RealFFT; validated on TEN and VoxConverse.
Akiki-AI/akiki-vad-1.0-int8

Audio Classification • 191k • Updated 16 days ago • 30

Note INT8 quantization of Akiki VAD 1.0. Unchanged FP32 RealFFT; validated on TEN and VoxConverse.
Akiki-AI/akiki-vad-1.0-int8-fp32-input

Audio Classification • 191k • Updated 16 days ago • 26

Note INT8-FP32-INPUT quantization of Akiki VAD 1.0. Unchanged FP32 RealFFT; validated on TEN and VoxConverse.
Akiki-AI/akiki-vad-1.0-int4

Audio Classification • 191k • Updated 16 days ago • 26

Note INT4 quantization of Akiki VAD 1.0. Unchanged FP32 RealFFT; validated on TEN and VoxConverse.
Akiki-AI/akiki-vad-silero623-v1-corpus-index

Viewer • Updated 16 days ago • 22 • 58

Note Training-data index and provenance. The full set of source datasets is linked from this index and the model card.