Fabio
c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
35 lines
1.3 KiB
Markdown
35 lines
1.3 KiB
Markdown
# AST AudioSet
|
|
|
|
AST AudioSet classifies an audio file into the 527 sound classes used by AudioSet. It uses the
|
|
[MIT `ast-finetuned-audioset-10-10-0.4593` checkpoint](https://huggingface.co/MIT/ast-finetuned-audioset-10-10-0.4593).
|
|
The default self-contained package is
|
|
`AST-AudioSet-GGUF/ast-audioset-f32.gguf` under
|
|
`audio-cpp/audio.cpp-gguf`. Only F32 is packaged.
|
|
|
|
```bash
|
|
build/debug/bin/audiocpp_cli \
|
|
--task cls \
|
|
--family ast_audioset \
|
|
--model /path/to/AST-AudioSet-GGUF/ast-audioset-f32.gguf \
|
|
--audio input.wav \
|
|
--backend cuda \
|
|
--threads 8 \
|
|
--request-option top_k=10 \
|
|
--log
|
|
```
|
|
|
|
## Request options (use with `--request-option`)
|
|
|
|
| Option | Values | Default | Description |
|
|
| --- | --- | --- | --- |
|
|
| `top_k` | `1` to `527` | `10` | Number of highest-scoring AudioSet classes to return. |
|
|
|
|
## Session options (use with `--session-option`)
|
|
|
|
| Option | Values | Default | Description |
|
|
| --- | --- | --- | --- |
|
|
| `ast_audioset.attention` | `auto`, `flash`, `eager` | `auto` | Attention lowering. |
|
|
| `ast_audioset.weight_type` | `native`, `f32`, `f16`, `bf16`, `q8_0`, `q4_0`, `q4_k` | `native` | Weight storage override. |
|
|
|
|
The model accepts ordinary audio inputs and resamples them to 16 kHz mono. Audio beyond the model's fixed
|
|
10.24-second feature window is truncated, matching the official feature extractor.
|