Add JIT-MPE and six comparison systems (three seeds each)

Original MIT-licensed GitHub release weights, calibration, model card and provenance for arXiv:2610.33113.

CITATION.cff

@@ -0,0 +1,26 @@
+cff-version: 1.2.0
+message: "If you use this code or the released weights, please cite the paper."
+title: "JIT-MPE: Just-in-Time Multi-Pitch Estimation under Output Deadlines"
+type: software
+authors:
+  - family-names: Kobayashi
+    given-names: Yoshiyuki
+    affiliation: Frieve
+repository-code: "https://github.com/Frieve-A/jit-mpe"
+license: MIT
+keywords:
+  - multi-pitch estimation
+  - real-time
+  - causal model
+  - music information retrieval
+preferred-citation:
+  type: article
+  title: "JIT-MPE: Just-in-Time Multi-Pitch Estimation under Output Deadlines"
+  authors:
+    - family-names: Kobayashi
+      given-names: Yoshiyuki
+      affiliation: Frieve
+  year: 2026
+  date-published: "2026-10-23"
+  doi: "10.48550/arXiv.2610.33113"
+  url: "https://arxiv.org/abs/2610.33113"

LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Yoshiyuki Kobayashi (Frieve-A)
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

README.md

@@ -0,0 +1,195 @@
+---
+license: mit
+library_name: pytorch
+tags:
+  - audio
+  - music
+  - multi-pitch-estimation
+  - music-information-retrieval
+  - causal
+  - arxiv:2610.33113
+---
+
+# JIT-MPE and CRNN comparison checkpoints
+
+Trained weights for the paper [JIT-MPE: Just-in-Time Multi-Pitch Estimation under Output Deadlines](https://arxiv.org/abs/2610.33113)
+(Yoshiyuki Kobayashi, Frieve, 2026).
+[Source code](https://github.com/Frieve-A/jit-mpe/tree/08bbd5acbb223962895f847a86c2d47a36179896) ·
+[Paper page](https://huggingface.co/papers/2610.33113) · [License](LICENSE) · [Data and software credits](THIRD_PARTY.md)
+
+JIT-MPE is a small causal model (about 86k parameters) that estimates which of the 88 piano pitches are sounding in
+mixed-ensemble audio. It processes the audio in 20 ms frames and can revise a frame until its output deadline
+(for example, 50 ms). On one CPU thread it takes about 2.4 ms per frame.
+
+This repository contains the same weights as the GitHub release, unchanged (no conversion or retraining). It also
+includes the comparison CRNN models that the author trained for the paper. The external systems compared in the
+paper are not included; get them from their own authors.
+
+## Models
+
+| System ID | Model | Parameters | Type |
+|---|---|---:|---|
+| `S` | JIT-MPE | 86,425 | Causal, five delay heads |
+| `A0` | Harmonic CRNN | 85,813 | Causal, one head |
+| `A1d` | Dilated CRNN | 95,767 | Causal, one head |
+| `B1` | Plain CRNN | 95,767 | Causal, one head |
+| `A0-bi` | Harmonic BiCRNN | 83,729 | Bidirectional, full-length input only |
+| `A1d-bi` | Dilated BiCRNN | 93,936 | Bidirectional, full-length input only |
+| `B1-bi` | Plain BiCRNN | 93,936 | Bidirectional, full-length input only |
+
+Each system was trained with seeds **1, 3 and 5**. These are three separate runs, not an ensemble; the paper reports
+their mean. Parameter counts exclude the auxiliary onset head, which is used only in training (it is still stored
+in the checkpoints).
+
+## Files
+
+```text
+weights/<SYSTEM>-s<SEED>.safetensors       # model weights (21 files)
+weights/<SYSTEM>-s<SEED>.calibration.json  # decision thresholds for each model (21 files)
+weights/head-selection-s<SEED>.json        # which delay head to use for each deadline (one per seed)
+configs/<SYSTEM>.json                      # training settings (not Transformers configs)
+requirements-paper.txt                     # Python environment used for the paper
+release-manifest.json                      # size and SHA-256 of every file, source revision, paper metadata
+LICENSE
+THIRD_PARTY.md
+CITATION.cff
+```
+
+To run one model you need three files: its `.safetensors`, its `.calibration.json` and the `head-selection` file for
+the same seed. Keep them in the same folder. The model code is in the GitHub package, so `AutoModel` and the
+Transformers `pipeline` cannot load these weights.
+
+## Usage
+
+Requirements: Python 3.10 or later and PyTorch for your platform. Install the source code at the matching revision
+and the Hugging Face Hub client:
+
+```bash
+git clone https://github.com/Frieve-A/jit-mpe.git
+cd jit-mpe
+git checkout 08bbd5acbb223962895f847a86c2d47a36179896
+python -m pip install -e .
+python -m pip install huggingface_hub==0.34.3
+```
+
+Python example (downloads only the three files for `S`, seed 1):
+
+```python
+from pathlib import Path
+import torch
+from huggingface_hub import snapshot_download
+from jit_mpe import load
+from jit_mpe.infer import estimate, read_audio, write
+
+name, seed = "S", 1
+snapshot = snapshot_download(
+    repo_id="Frieve-A/jit-mpe",
+    revision="main",  # replace with a full Hub commit hash to pin a published snapshot
+    allow_patterns=[
+        f"weights/{name}-s{seed}.safetensors",
+        f"weights/{name}-s{seed}.calibration.json",
+        f"weights/head-selection-s{seed}.json",
+    ],
+)
+system = load(name, seed, weights_dir=Path(snapshot) / "weights", device="cpu")
+torch.set_num_threads(1)
+with torch.inference_mode():
+    times, frequencies = estimate(
+        read_audio("song.wav"), system,
+        clock="scheduled_stale_hold", deadline_ms=50,
+    )
+write("song.csv", times, frequencies)
+```
+
+Command-line example (replace `HUB_COMMIT` with a commit hash of this repository, or use `main`):
+
+```bash
+hf download Frieve-A/jit-mpe weights/S-s1.safetensors weights/S-s1.calibration.json weights/head-selection-s1.json --revision HUB_COMMIT --local-dir hub-model
+python -m jit_mpe.infer song.wav --weights hub-model/weights --system S --seed 1 --clock scheduled_stale_hold --deadline 50 --out song.csv
+```
+
+### Input and output
+
+- **Input:** any audio file. `read_audio` mixes it down to mono and resamples it to 44.1 kHz.
+- **Output:** one row every 10 ms. A CSV row has the time in seconds, followed by the frequencies (Hz) of the pitches
+  that are sounding. An NPZ file has `times [Q]` and `frequencies [Q, 88]`; absent pitches are 0.
+- **Pitch range:** MIDI 21–108 (A0–C8), with a fine frequency estimate for each active pitch.
+
+### Deadlines
+
+A deadline D means that the answer for time q must be ready by q + D. With `scheduled_stale_hold`, the output for q
+is the most recent frame that is ready by then, so there is always an answer, but it may be slightly old.
+
+`S` has five delay heads that look 0, 20, 40, 80 and 160 ms ahead. For each deadline, the head-selection file says
+which head to use; for 50 ms it is the 40 ms head. To process a whole file without a deadline, omit `clock` and
+`deadline_ms` (Python) or `--clock` and `--deadline` (CLI). The `-bi` models only work this way, because they use
+future audio.
+
+`estimate` reads a complete file and reproduces what a real-time system would have output by each deadline. It is
+not a live audio callback; for frame-by-frame streaming with `model.step`, see `bench/cpu.py` in the source code.
+
+## Training
+
+The models were trained on GuitarSet, Slakh2100-redux, MusicNet, Schubert Winterreise, Cantoria, Dagstuhl ChoirSet,
+vocadito and Groove MIDI. Versions, links and licenses are listed in [THIRD_PARTY.md](THIRD_PARTY.md). This repository
+does not include any training audio, annotations, prepared features or external models.
+
+Training has two stages: 10,400 steps at learning rate 0.002, then 3,000 steps at 0.0005 with an added onset loss
+(batch size 16, 192-frame clips, 16 warm-up frames). Afterwards, the decision thresholds are fitted on a calibration
+split and the delay heads are selected on a development set. The GitHub README gives the full commands for data
+preparation, training and evaluation. `requirements-paper.txt` records the environment used for the paper
+(Python 3.12, PyTorch 2.5.1); a CUDA build is not needed for CPU inference.
+
+## Results
+
+The numbers below are copied from Table 4 of the paper ([arXiv v1](https://arxiv.org/html/2610.33113v1)); they were
+not re-measured for this upload. They are F1 scores (%) for JIT-MPE with `scheduled_stale_hold` at a 50 ms deadline,
+averaged per recording and over the three seeds, with pitches rounded to semitones.
+
+| Held-out set | F1 (%) |
+|---|---:|
+| AAM | 82.6 |
+| ChoraleBricks | 78.9 |
+| BSED-real | 64.5 |
+| BSED-synthetic | 77.0 |
+
+Notes:
+
+- The paper also reports 50-cent (continuous f0) scores and full-length scores. These use different settings and
+  should not be compared directly with the table above.
+- ChoraleBricks is annotated with continuous f0; the semitone score above is given only so that all sets use the
+  same metric.
+- The BSED recordings were not used in training, but some of the pieces also appear in the MusicNet training data.
+
+See the paper for development-set results, comparisons with other systems, confidence intervals and scoring details.
+
+## Intended use and limitations
+
+Intended for research on frame-level pitch estimation, pitch-display prototypes and experiments with output
+deadlines. The models do not identify instruments and do not produce note events (onsets/offsets).
+
+- The paper does not evaluate note-level transcription, singing voice or real popular-music recordings.
+- Performance on other instruments or recording conditions has not been tested.
+- The reported CPU timings depend on the hardware and do not include audio I/O, scheduling or display latency.
+- Retraining with the released code will not give bit-identical weights, because these weights were trained before
+  deterministic training was added. The public vocadito split also differs from the one used in the paper (see the
+  determinism section of the GitHub README). To reproduce the paper's inference results, use these checkpoints.
+
+## License and citation
+
+The code and these weights are released by the author under the [MIT license](LICENSE). This does not cover the
+training datasets or external models, which keep their own licenses. In particular, the Schubert Winterreise
+recordings have their own conditions, described in the original dataset release.
+
+```bibtex
+@misc{kobayashi2026jitmpe,
+  title = {{JIT-MPE}: Just-in-Time Multi-Pitch Estimation under Output Deadlines},
+  author = {Kobayashi, Yoshiyuki},
+  year = {2026},
+  eprint = {2610.33113},
+  archivePrefix = {arXiv},
+  primaryClass = {cs.SD},
+  doi = {10.48550/arXiv.2610.33113},
+  url = {https://arxiv.org/abs/2610.33113}
+}
+```

THIRD_PARTY.md

@@ -0,0 +1,59 @@
+# Third-party data and software
+
+This repository contains no third-party audio, annotations or model code. The datasets and external systems below are
+obtained from their distributors by the user; the scripts here only read them from paths given on the command line.
+Each dataset and system is subject to its own licence and terms. The code and the trained weights in this repository are
+released under the MIT licence (see `LICENSE`).
+
+## Training data
+
+All seven model families are trained on clips from the following datasets. Clip extraction and the train/validation
+fold assignment are seeded rules in `jit_mpe/data/packs.py`; no list of files is distributed.
+
+| Dataset | Source | Licence (as stated by the distributor) |
+|---|---|---|
+| GuitarSet v1.1.0 | https://zenodo.org/records/3371780 | CC BY 4.0 |
+| Slakh2100-redux | https://zenodo.org/records/4599666 | CC BY 4.0 (rendered from Lakh MIDI v0.1, CC BY 4.0) |
+| MusicNet | https://zenodo.org/records/5120004 | CC BY 4.0 |
+| Schubert Winterreise Dataset v2.1 | https://zenodo.org/records/10839767 | CC BY 3.0; individual recordings carry further terms (see its README) |
+| Cantoria v1.0.0 | https://zenodo.org/records/5878677 | CC BY 4.0 |
+| Dagstuhl ChoirSet v1.2.3 | https://zenodo.org/records/4618287 | CC BY 4.0 |
+| vocadito v3 | https://zenodo.org/records/5578807 | CC BY 4.0 |
+| Groove MIDI Dataset v1.0.0 | https://magenta.tensorflow.org/datasets/groove | CC BY 4.0 |
+
+## Evaluation data (evaluation only)
+
+| Dataset | Source | Licence (as stated by the distributor) |
+|---|---|---|
+| URMP, 16 kHz release prepared for YourMT3 (`urmp_yourmt3_16k.tar.gz`, MD5 `4f539c71678a77ba34f6dfca41072102`) | https://zenodo.org/records/10009959 (original dataset: https://labsites.rochester.edu/air/projects/URMP.html) | CC BY 4.0 (record); the original URMP states no licence |
+| Bach10-mf0-synth v1.0.0 | https://zenodo.org/records/1481156 | CC BY-NC 4.0 |
+| MusicNet (development recordings) | https://zenodo.org/records/5120004 | CC BY 4.0 |
+| TRIOS | https://zenodo.org/records/6797837 | CC BY-NC-SA 2.0 UK |
+| PHENICX-Anechoic | https://zenodo.org/records/840025 | annotations CC BY-NC-SA 3.0, non-commercial use, no redistribution |
+| AAM v1.1.0 | https://zenodo.org/records/5794629 | CC BY 4.0 |
+| ChoraleBricks v1.1.0 | https://zenodo.org/records/20849469 | CC BY 4.0 |
+| BSED v1.0.0 | https://zenodo.org/records/20344500 | CC BY-NC-SA 4.0 |
+
+## External systems (compared in the paper)
+
+The scripts in `external/` fetch each system from its upstream source at the pinned revision and run it with the small
+runtime adaptations listed in `external/README.md`; the upstream files are not edited and nothing is vendored. Their
+extra Python requirements are listed in `requirements-external.txt`.
+
+| System | Upstream | Revision | Licence of the code |
+|---|---|---|---|
+| SS-MPE v0.1.0, v0.2.0 | https://github.com/cwitkowitz/ss-mpe | releases v0.1.0, v0.2.0 | MIT |
+| Timbre-Trap | https://github.com/sony/timbre-trap (weights: `models/tt-orig.pt` of the Hugging Face Space `cwitkowitz/timbre-trap`) | code `7afe7e9b327929c099baeccd4b21973aedb94d9b`, Space `84fbd38582435c863a2d65197a75edd794888f19` | MIT (code); weights: see upstream |
+| Basic Pitch | https://github.com/spotify/basic-pitch | `fa5997af0a8210982619003269994a1be25eddf3` | Apache 2.0 |
+| WP-DRCNN, WP-PUnet XL | https://github.com/christofw/multipitch_architectures | `7e9c1bcb83e311c370cdfffe599d645d1ebae640` | CC0 1.0 |
+| YourMT3+ | Hugging Face Space `mimbres/YourMT3` (code and checkpoint; project repository https://github.com/mimbres/YourMT3) | Space `5e66c1ea173a8186e0d20432b841d3180cc015b5` | see upstream |
+| MuScriptor | https://github.com/muscriptor/muscriptor (weights: Hugging Face `MuScriptor/muscriptor-large`, gated; requires CUDA) | code `7f213afecf23bd6a1b8672aa223690ee9807cefb`, weights `8809fdfbed2affa7ade94a7059e746e3880720e7` | MIT (code); weights: see upstream |
+
+The checkpoints of the external systems are distributed by their authors under their own terms.
+
+## Python dependencies
+
+PyTorch (BSD-3-Clause), NumPy (BSD-3-Clause), SciPy (BSD-3-Clause), safetensors (Apache 2.0), soundfile (BSD-3-Clause),
+soxr (LGPL-2.1), numba (BSD-2-Clause), mido (MIT), pretty_midi (MIT), PyYAML (MIT), matplotlib (PSF-based);
+for `external/`, the packages in `requirements-external.txt` under their own licences. They are installed from PyPI
+and are not included in this repository.

configs/A0-bi.json

@@ -0,0 +1,10 @@
+{
+  "system": "A0-bi",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/A0.json

@@ -0,0 +1,10 @@
+{
+  "system": "A0",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/A1d-bi.json

@@ -0,0 +1,10 @@
+{
+  "system": "A1d-bi",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/A1d.json

@@ -0,0 +1,10 @@
+{
+  "system": "A1d",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/B1-bi.json

@@ -0,0 +1,10 @@
+{
+  "system": "B1-bi",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/B1.json

@@ -0,0 +1,10 @@
+{
+  "system": "B1",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

configs/S.json

@@ -0,0 +1,10 @@
+{
+  "system": "S",
+  "packs": ["train", "vocal"],
+  "weights": {"groove": 0.3},
+  "B": 16,
+  "T": 192,
+  "burn": 16,
+  "stage1": {"steps": 10400, "lr": 0.002},
+  "stage2": {"steps": 3000, "lr": 0.0005, "onset_weight": 0.18, "onset_pos_weight": 10.0, "onset_J": 2.0}
+}

release-manifest.json

@@ -0,0 +1,265 @@
+{
+  "files": {
+    "CITATION.cff": {
+      "sha256": "ae958c86ed5a5c21d1b34b63922797343fac5d6fdc6b4c54ccaca249ffc7631f",
+      "size": 779
+    },
+    "LICENSE": {
+      "sha256": "f286d8a2d9ddde223d3aba5c6a689e662d65616d52aa619a81ea86c8f0f52753",
+      "size": 1087
+    },
+    "README.md": {
+      "sha256": "7739fd3adba5a0bbfa4567b03ef8849db841605b3fe1594a7c808642ed464410",
+      "size": 9159
+    },
+    "THIRD_PARTY.md": {
+      "sha256": "bb0b57a698d572aad6483dede2665646da8d6aa98759af10f97517ea7b02a87a",
+      "size": 4428
+    },
+    "configs/A0-bi.json": {
+      "sha256": "21d55933d03798feca6b8e06908c447043d631960c9ae0e8d51a9f547492f316",
+      "size": 272
+    },
+    "configs/A0.json": {
+      "sha256": "9c5e64f378c7253a787b8f3fe92e6e9d5ba1738c2b1a50efd53186d6b938f9f9",
+      "size": 269
+    },
+    "configs/A1d-bi.json": {
+      "sha256": "392ad2669f35c3d5223a8185bb7a57e8082639da55198439d4e80378e0f447cf",
+      "size": 273
+    },
+    "configs/A1d.json": {
+      "sha256": "de37630394c9e55c53af31e1a956a9655e758463ac56a3efc4cded322ca576b2",
+      "size": 270
+    },
+    "configs/B1-bi.json": {
+      "sha256": "e802cb7e3efd3021bf3b94fe8c649a785003c56b2ae074d9b225618cd39629a3",
+      "size": 272
+    },
+    "configs/B1.json": {
+      "sha256": "2c023f0a6f4e399d4517586cae643f1be2baf3d83c5f64a6c248a31984aaca0f",
+      "size": 269
+    },
+    "configs/S.json": {
+      "sha256": "d81bcc68b98c41343a0c74cf6e78853d7578aa06e2d59338fadef7a40a790134",
+      "size": 268
+    },
+    "requirements-paper.txt": {
+      "sha256": "9682df96feb43bdd8edf22acbcf4c524a6016117678fcc1f8104cdc13ac9e3bb",
+      "size": 283
+    },
+    "weights/A0-bi-s1.calibration.json": {
+      "sha256": "82136273e46cb4e90b8a2f58d1b1d736fafb061c0991c251ebe098cc5eb9b72a",
+      "size": 174
+    },
+    "weights/A0-bi-s1.safetensors": {
+      "sha256": "789ddc5ebdbcced0347362358947d7d82fb380260795077935ca362b71350a4a",
+      "size": 423916
+    },
+    "weights/A0-bi-s3.calibration.json": {
+      "sha256": "b64b59dc593a691a8471928985e506fef48f83e93d7ce9d31689b5a63e115746",
+      "size": 176
+    },
+    "weights/A0-bi-s3.safetensors": {
+      "sha256": "3ec02530f3b619e113381ae5e15345905c3781854c46a9d81af5c39f696d561b",
+      "size": 423916
+    },
+    "weights/A0-bi-s5.calibration.json": {
+      "sha256": "e1f345a5886bd7bbf17d2a259c2c95e4fe9c070e59a2a2abef4c90e584e5f64e",
+      "size": 172
+    },
+    "weights/A0-bi-s5.safetensors": {
+      "sha256": "a0e64cb05b82a44616be6b2145904992e980bda4f47fe279c6e22cb08abeb52c",
+      "size": 423916
+    },
+    "weights/A0-s1.calibration.json": {
+      "sha256": "bf6c962288b6611f981f244fa5ce0f20c1b2fb4785633ae8313b728b4e528eb2",
+      "size": 172
+    },
+    "weights/A0-s1.safetensors": {
+      "sha256": "9e961901424bdae2547c8233c45dacc6815d186b3996752cc6a81ee25523cf9c",
+      "size": 421332
+    },
+    "weights/A0-s3.calibration.json": {
+      "sha256": "6f673230122cbf569c58b5d00cb42f78c39bdf44654bcee534e21eca4a5d97d9",
+      "size": 172
+    },
+    "weights/A0-s3.safetensors": {
+      "sha256": "721cfe88b75c165019eaad7b8e32fe4ecf519288ee5414f928fef0b9580a2264",
+      "size": 421332
+    },
+    "weights/A0-s5.calibration.json": {
+      "sha256": "23639b759a495e99aae6d9586fb991f63fdf0a04c39aeebbdfde41b67c426b49",
+      "size": 175
+    },
+    "weights/A0-s5.safetensors": {
+      "sha256": "3a43537ca8d892962d652686a345f044d250943b145cc3a6a9c94215d4370c4f",
+      "size": 421332
+    },
+    "weights/A1d-bi-s1.calibration.json": {
+      "sha256": "3a6e1d9a6c9dec3200ee89dfb307b6cf24b526a08c62e657d128281bcf2fa3f6",
+      "size": 174
+    },
+    "weights/A1d-bi-s1.safetensors": {
+      "sha256": "2e75301f5bcde4b56f6ec5cac1bd24a8b3be14766ec2db208a87b4e4d039618b",
+      "size": 438416
+    },
+    "weights/A1d-bi-s3.calibration.json": {
+      "sha256": "893cdf343885137640909e6fae3298048ecc8898e3a25906521ec24ed5198e59",
+      "size": 175
+    },
+    "weights/A1d-bi-s3.safetensors": {
+      "sha256": "b11ade2a2e2a353679cd151b2877fc3d4d5e606bfaf4b495a90c5936db42f331",
+      "size": 438416
+    },
+    "weights/A1d-bi-s5.calibration.json": {
+      "sha256": "aea27416f57b07da4a87673ab0c81490143a9d200f56e8d846e5b5e4f32f4c08",
+      "size": 174
+    },
+    "weights/A1d-bi-s5.safetensors": {
+      "sha256": "3116157db00241c643eaabf3dbb8b1a1577164e36c473fe4d6b201903827eab3",
+      "size": 438416
+    },
+    "weights/A1d-s1.calibration.json": {
+      "sha256": "189e07bdc5cdcfe8fbc10cf365694dcb2b9666e474462746c5e97cbcc2f34112",
+      "size": 173
+    },
+    "weights/A1d-s1.safetensors": {
+      "sha256": "26e33c5cdafe3da406ecbceb8d8b4524111ea01e285a2a641237e01e205b6c58",
+      "size": 440116
+    },
+    "weights/A1d-s3.calibration.json": {
+      "sha256": "97e5be31cf1463b0b595eabc11e6dbe35deee2ea5fa0fbc7ac8321f75880b8e0",
+      "size": 176
+    },
+    "weights/A1d-s3.safetensors": {
+      "sha256": "92bd775de51254d50a2fcfa000a552209a98f853a56031e02db0c0275fb7d095",
+      "size": 440116
+    },
+    "weights/A1d-s5.calibration.json": {
+      "sha256": "fb326a9e74fa605c24af20027798f9233a4fa73cc7f2c8ab220f201522f53780",
+      "size": 174
+    },
+    "weights/A1d-s5.safetensors": {
+      "sha256": "0b611800820c0285bdbe10c4707cf9c7983882c4d993cdef8dbe9ae43d71248c",
+      "size": 440116
+    },
+    "weights/B1-bi-s1.calibration.json": {
+      "sha256": "ce1ba35cd60e68d22a68d8c2a1b0ee63e6772a5f44eaa8d31f5d5c3db0dae3e8",
+      "size": 174
+    },
+    "weights/B1-bi-s1.safetensors": {
+      "sha256": "4bb19091a42e768472ffad775f49c36b49d709b7429fae645fbf297792be7ffb",
+      "size": 438416
+    },
+    "weights/B1-bi-s3.calibration.json": {
+      "sha256": "89b2912de82c44b285ba29be6e35dfc9126d487e52d4ac4d20cb352736bb7f4b",
+      "size": 175
+    },
+    "weights/B1-bi-s3.safetensors": {
+      "sha256": "7842a11b23c186b1b46ffb7cb1a2b0d550203dc302c055e41dfc25a1276e77a0",
+      "size": 438416
+    },
+    "weights/B1-bi-s5.calibration.json": {
+      "sha256": "a1c70164734821a414de6dbf74bfce8e44fa5f86a3f4817577fafc41a1c4810f",
+      "size": 176
+    },
+    "weights/B1-bi-s5.safetensors": {
+      "sha256": "bbf4b8e9868ab45085372c86e0fa8632a5934b4c0a5b5e00074201bf6134c0ff",
+      "size": 438416
+    },
+    "weights/B1-s1.calibration.json": {
+      "sha256": "b394627c3cb8eb0b794bea241d7a6cd99e6a79c7e864b5aee7af89a1015dd51b",
+      "size": 176
+    },
+    "weights/B1-s1.safetensors": {
+      "sha256": "15aeff68451495100137de148a282b25d6ba6b7f84da0148ed2fab3b3d04217d",
+      "size": 440116
+    },
+    "weights/B1-s3.calibration.json": {
+      "sha256": "ac0330f4aace73e8ea8668ba8647be4838b85a0b486048b1a6461bbf6710837e",
+      "size": 175
+    },
+    "weights/B1-s3.safetensors": {
+      "sha256": "797dcf0da012e1d3eba2fb976a98defa38398738d211c2621a6d83528ba768c5",
+      "size": 440116
+    },
+    "weights/B1-s5.calibration.json": {
+      "sha256": "584aba136a0d5e4d8a82a6d7bc0816aa02fa36286c22079ea4c32f2a1aa2030f",
+      "size": 174
+    },
+    "weights/B1-s5.safetensors": {
+      "sha256": "375c7edad20162d8e9c58be822263cdb1532b94dad1c4ba880e4af32397acf1b",
+      "size": 440116
+    },
+    "weights/S-s1.calibration.json": {
+      "sha256": "13375e7be57c5e7fe26e22fb6829daa101fdc013645a9bc1fdade95bd069b749",
+      "size": 379
+    },
+    "weights/S-s1.safetensors": {
+      "sha256": "54c053610496ccc757551f718de97f70438b7c81044e71bba6e3e1be52cf4ae8",
+      "size": 423780
+    },
+    "weights/S-s3.calibration.json": {
+      "sha256": "6d27ff8160b6feba7770881304ab5c567e169b39ac7593528299be6282341311",
+      "size": 379
+    },
+    "weights/S-s3.safetensors": {
+      "sha256": "5cda819af6ce132d866ad8ad8671af857fe1ea2b322f3a7f810e86495fbbfbea",
+      "size": 423780
+    },
+    "weights/S-s5.calibration.json": {
+      "sha256": "3ea4bd8bbf4b7a2308ce10446f66fdc6a8bbfd01809b9ac4b813befe8bc03b30",
+      "size": 374
+    },
+    "weights/S-s5.safetensors": {
+      "sha256": "6b5cebf336ea9f5ed7d8f0184c3ecdba0869f65a35fa91d26edd84bec11f9e77",
+      "size": 423780
+    },
+    "weights/head-selection-s1.json": {
+      "sha256": "ddaf44d54ec17c502ba10a12e1ab16e691741f504c3d628a4e139e6add67cee9",
+      "size": 2238
+    },
+    "weights/head-selection-s3.json": {
+      "sha256": "62eb6c8ade5594aecc5d1c5885f62907234ebaef26fe8d91d5fb17416ff4618b",
+      "size": 2238
+    },
+    "weights/head-selection-s5.json": {
+      "sha256": "698cecb09e94c40176477f3380557b77f67eaae2457b2d9ea6c2ac599de01bc3",
+      "size": 2238
+    }
+  },
+  "license": "mit",
+  "paper": {
+    "affiliation": "Frieve",
+    "arxiv_id": "2610.33113",
+    "arxiv_version": "v1",
+    "authors": [
+      "Yoshiyuki Kobayashi"
+    ],
+    "doi": "10.48550/arXiv.2610.33113",
+    "published": "2026-10-23",
+    "title": "JIT-MPE: Just-in-Time Multi-Pitch Estimation under Output Deadlines",
+    "url": "https://arxiv.org/abs/2610.33113"
+  },
+  "repo_id": "Frieve-A/jit-mpe",
+  "schema_version": 1,
+  "seeds": [
+    1,
+    3,
+    5
+  ],
+  "source": {
+    "repository": "https://github.com/Frieve-A/jit-mpe",
+    "revision": "08bbd5acbb223962895f847a86c2d47a36179896"
+  },
+  "systems": [
+    "S",
+    "A0",
+    "A1d",
+    "B1",
+    "A0-bi",
+    "A1d-bi",
+    "B1-bi"
+  ]
+}

requirements-paper.txt

@@ -0,0 +1,13 @@
+# Versions used to produce the results in the paper (Python 3.12, CUDA 12.4 build of torch).
+torch==2.5.1
+numpy==2.2.4
+scipy==1.15.2
+safetensors==0.5.3
+soundfile==0.13.1
+soxr==0.5.0.post1
+numba==0.61.2
+mido==1.3.3
+pretty_midi==0.2.11.post0
+pyyaml==6.0.2
+matplotlib==3.10.1
+pytest>=8

weights/A0-bi-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.35
+  ],
+  "registers": {
+    "21-47": [
+      0.45
+    ],
+    "48-71": [
+      0.3
+    ],
+    "72-91": [
+      0.5
+    ]
+  }
+}

weights/A0-bi-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:789ddc5ebdbcced0347362358947d7d82fb380260795077935ca362b71350a4a
+size 423916

weights/A0-bi-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.35
+  ],
+  "registers": {
+    "21-47": [
+      0.65
+    ],
+    "48-71": [
+      0.35
+    ],
+    "72-91": [
+      0.85
+    ]
+  }
+}

weights/A0-bi-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ec02530f3b619e113381ae5e15345905c3781854c46a9d81af5c39f696d561b
+size 423916

weights/A0-bi-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.4
+  ],
+  "registers": {
+    "21-47": [
+      0.4
+    ],
+    "48-71": [
+      0.4
+    ],
+    "72-91": [
+      0.9
+    ]
+  }
+}

weights/A0-bi-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0e64cb05b82a44616be6b2145904992e980bda4f47fe279c6e22cb08abeb52c
+size 423916

weights/A0-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.3
+  ],
+  "registers": {
+    "21-47": [
+      0.3
+    ],
+    "48-71": [
+      0.3
+    ],
+    "72-91": [
+      0.5
+    ]
+  }
+}

weights/A0-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e961901424bdae2547c8233c45dacc6815d186b3996752cc6a81ee25523cf9c
+size 421332

weights/A0-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.4
+  ],
+  "registers": {
+    "21-47": [
+      0.5
+    ],
+    "48-71": [
+      0.4
+    ],
+    "72-91": [
+      0.6
+    ]
+  }
+}

weights/A0-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:721cfe88b75c165019eaad7b8e32fe4ecf519288ee5414f928fef0b9580a2264
+size 421332

weights/A0-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.45
+  ],
+  "registers": {
+    "21-47": [
+      0.6
+    ],
+    "48-71": [
+      0.45
+    ],
+    "72-91": [
+      0.55
+    ]
+  }
+}

weights/A0-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a43537ca8d892962d652686a345f044d250943b145cc3a6a9c94215d4370c4f
+size 421332

weights/A1d-bi-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.4
+  ],
+  "registers": {
+    "21-47": [
+      0.45
+    ],
+    "48-71": [
+      0.3
+    ],
+    "72-91": [
+      0.75
+    ]
+  }
+}

weights/A1d-bi-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e75301f5bcde4b56f6ec5cac1bd24a8b3be14766ec2db208a87b4e4d039618b
+size 438416

weights/A1d-bi-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.55
+  ],
+  "registers": {
+    "21-47": [
+      0.65
+    ],
+    "48-71": [
+      0.55
+    ],
+    "72-91": [
+      0.7
+    ]
+  }
+}

weights/A1d-bi-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b11ade2a2e2a353679cd151b2877fc3d4d5e606bfaf4b495a90c5936db42f331
+size 438416

weights/A1d-bi-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.3
+  ],
+  "registers": {
+    "21-47": [
+      0.25
+    ],
+    "48-71": [
+      0.3
+    ],
+    "72-91": [
+      0.75
+    ]
+  }
+}

weights/A1d-bi-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3116157db00241c643eaabf3dbb8b1a1577164e36c473fe4d6b201903827eab3
+size 438416

weights/A1d-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.3
+  ],
+  "registers": {
+    "21-47": [
+      0.35
+    ],
+    "48-71": [
+      0.3
+    ],
+    "72-91": [
+      0.6
+    ]
+  }
+}

weights/A1d-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26e33c5cdafe3da406ecbceb8d8b4524111ea01e285a2a641237e01e205b6c58
+size 440116

weights/A1d-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.55
+  ],
+  "registers": {
+    "21-47": [
+      0.55
+    ],
+    "48-71": [
+      0.55
+    ],
+    "72-91": [
+      0.65
+    ]
+  }
+}

weights/A1d-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92bd775de51254d50a2fcfa000a552209a98f853a56031e02db0c0275fb7d095
+size 440116

weights/A1d-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.45
+  ],
+  "registers": {
+    "21-47": [
+      0.5
+    ],
+    "48-71": [
+      0.45
+    ],
+    "72-91": [
+      0.6
+    ]
+  }
+}

weights/A1d-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b611800820c0285bdbe10c4707cf9c7983882c4d993cdef8dbe9ae43d71248c
+size 440116

weights/B1-bi-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.55
+  ],
+  "registers": {
+    "21-47": [
+      1.2
+    ],
+    "48-71": [
+      0.4
+    ],
+    "72-91": [
+      0.65
+    ]
+  }
+}

weights/B1-bi-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bb19091a42e768472ffad775f49c36b49d709b7429fae645fbf297792be7ffb
+size 438416

weights/B1-bi-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    -0.2
+  ],
+  "registers": {
+    "21-47": [
+      0.6
+    ],
+    "48-71": [
+      -0.2
+    ],
+    "72-91": [
+      0.15
+    ]
+  }
+}

weights/B1-bi-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7842a11b23c186b1b46ffb7cb1a2b0d550203dc302c055e41dfc25a1276e77a0
+size 438416

weights/B1-bi-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.75
+  ],
+  "registers": {
+    "21-47": [
+      1.45
+    ],
+    "48-71": [
+      0.65
+    ],
+    "72-91": [
+      1.15
+    ]
+  }
+}

weights/B1-bi-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbf4b8e9868ab45085372c86e0fa8632a5934b4c0a5b5e00074201bf6134c0ff
+size 438416

weights/B1-s1.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.55
+  ],
+  "registers": {
+    "21-47": [
+      1.15
+    ],
+    "48-71": [
+      0.55
+    ],
+    "72-91": [
+      0.65
+    ]
+  }
+}

weights/B1-s1.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15aeff68451495100137de148a282b25d6ba6b7f84da0148ed2fab3b3d04217d
+size 440116

weights/B1-s3.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.65
+  ],
+  "registers": {
+    "21-47": [
+      1.15
+    ],
+    "48-71": [
+      0.65
+    ],
+    "72-91": [
+      0.7
+    ]
+  }
+}

weights/B1-s3.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:797dcf0da012e1d3eba2fb976a98defa38398738d211c2621a6d83528ba768c5
+size 440116

weights/B1-s5.calibration.json

@@ -0,0 +1,19 @@
+{
+  "delays": [
+    0
+  ],
+  "global": [
+    0.65
+  ],
+  "registers": {
+    "21-47": [
+      1.1
+    ],
+    "48-71": [
+      0.65
+    ],
+    "72-91": [
+      0.6
+    ]
+  }
+}

weights/B1-s5.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:375c7edad20162d8e9c58be822263cdb1532b94dad1c4ba880e4af32397acf1b
+size 440116

weights/S-s1.calibration.json

@@ -0,0 +1,39 @@
+{
+  "delays": [
+    0,
+    1,
+    2,
+    4,
+    8
+  ],
+  "global": [
+    0.15,
+    0.25,
+    0.25,
+    0.2,
+    0.25
+  ],
+  "registers": {
+    "21-47": [
+      0.6,
+      0.4,
+      0.25,
+      0.35,
+      0.3
+    ],
+    "48-71": [
+      0.15,
+      0.25,
+      0.2,
+      0.2,
+      0.25
+    ],
+    "72-91": [
+      0.35,
+      0.3,
+      0.35,
+      0.2,
+      0.3
+    ]
+  }
+}