Modelpublic

cstr/basic-pitch-GGUF

sourceHugging Faceapache-2.0updated 24d agoView on Hugging Face

0likes360downloads

Model Card

Basic Pitch — GGUF (ggml)

GGUF / ggml conversion of Spotify Basic Pitch (ICASSP 2022 nmp model, Apache-2.0) for [CrispStrobe/CrispASR](https://github.com/CrispStrobe/CrispASR) (§250): lightweight polyphonic audio → note-event transcription (any instrument, 88-key MIDI range).

Converted from the nmp.onnx that ships inside the basic-pitch repo (sha256 2c3c1d14…, 230 KB) with `models/convert-basic-pitch-to-gguf.py`. The nnAudio CQT2010v2 kernels, the 256-tap decimation FIR, and the rescale vector are copied bit-for-bit from the ONNX initializers — nothing DSP-critical is reimplemented.

Files

File	Size	Notes
`basic-pitch-f32.gguf`	146 KB	exact — note events match the Python reference 27/27 and 11/11 on the validation clips
`basic-pitch-f16.gguf`	113 KB	default; two note ends may shift by one frame (11.6 ms) at threshold boundaries

Usage

bash

crispasr --piano --backend basic-pitch -m auto --auto-download -f input.wav
crispasr --piano -m basic-pitch-f32.gguf -f input.wav --piano-format json

Parity vs the upstream ONNX (per-stage cosine ≥ 0.9991 on real audio, 1.000000 on native-rate input; end-to-end note events exact at F32) is enforced by the crispasr-diff basic-pitch harness with tools/reference_backends/basic_pitch.py as the oracle.