| """Kaggle notebook body: release-candidate run at the operating point. |
| |
| 4000 steps was the best configuration; 8000 regressed. This run reproduces it |
| with the teacher seeded, and this time keeps the audio so the result can be |
| judged by ear rather than only by CER. |
| |
| Produces: |
| train/student4.pt weights |
| train/dit4.onnx 412 MB fp32 |
| train/dit4_int4.onnx 70 MB int4 |
| train/eval_rows*.json teacher / student / int4 on a07 + a08 |
| train/eval_wavs/*.wav teacher and student audio for the same cells |
| |
| Uses the cached trajectories, so no regeneration. |
| """ |
| import os |
| import subprocess |
| import sys |
|
|
| from notebook_eval import BUNDLE, MODEL, ROOT, WEIGHTS, pull, stage_deps, stage_repo |
|
|
| BUNDLE_REPO = "idefinity/webfm-eval" |
| HOLDOUT = "a07" |
| EVAL_VOICES = "a07,a08" |
| TRAIN_STEPS = "4000" |
|
|
|
|
| def run(tag: str, extra: dict, gpu: int = 0) -> int: |
| env = dict(os.environ, WORK_ROOT=str(ROOT), HOLDOUT=HOLDOUT, **extra) |
| env["CUDA_VISIBLE_DEVICES"] = str(gpu) |
| env["PYTORCH_CUDA_ALLOC_CONF"] = "expandable_segments:True" |
| log = open(ROOT / f"final_{tag}.log", "w") |
| p = subprocess.Popen([sys.executable, str(BUNDLE / "kaggle/train_voices.py")], |
| env=env, stdout=log, stderr=subprocess.STDOUT, cwd=str(ROOT)) |
| rc = p.wait() |
| print(f"[{tag}] rc={rc}", flush=True) |
| if rc != 0: |
| print((ROOT / f"final_{tag}.log").read_text()[-2500:]) |
| return rc |
|
|
|
|
| def main() -> None: |
| stage_repo() |
| stage_deps() |
| pull(WEIGHTS, MODEL) |
| pull(BUNDLE_REPO, BUNDLE, repo_type="dataset") |
| if not (BUNDLE / "traj.zip").exists(): |
| raise SystemExit("traj.zip missing") |
|
|
| if run("main", {"STAGE": "all", "TRAIN_STEPS": TRAIN_STEPS, "BATCH": "16", |
| "EVAL_VOICES": EVAL_VOICES, "SAVE_WAVS": "1"}) != 0: |
| raise SystemExit("train/export/eval failed") |
|
|
| run("int4", {"STAGE": "eval_int4", "EVAL_VOICES": EVAL_VOICES}) |
|
|
|
|
| if __name__ == "__main__": |
| main() |
|
|
