Docker-сервис, выбор основы и сравнение моделей, сборщик датасета
docker / build (push) Failing after 39s

- Docker: Dockerfile (CPU), compose с FlareSolverr, настройка через DUB_* переменные,
  токен доступа к /predict, tools/pack_models.py, workflow Gitea Actions, docs/docker.md
- Модель грузится офлайн (config.json рядом с весами); общий код вынесен в src/spans.py,
  сервису не нужны sklearn/matplotlib/pandas
- Обучение: профили tiny2/base/large/modern, запуски в models/runs, реестр и честное
  сравнение на одной тестовой выборке (src/runs.py), пункт меню «Модели»
- Датасет: язык и аудио-параметры дорожек, 18+ по строкам «Аудио», ручные метки по ссылкам,
  переиндексация сохранённого HTML, параллельная сборка, сбалансированная выборка
- Тексты страниц (data/pages) больше не хранятся в git

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
This commit is contained in:
DenozordecandClaude Sonnet 5.5 committed 2026-10-28 17:19:10 +07:00
1 parent d155b8584b
commit efc9a4f56a
411 files changed
+87495 -27363

No files matched your search

+82 -62
View File
@@ -2,120 +2,140 @@
"history": [
{
"epoch": 0,
"loss": 2.357788513909589,
"type_accuracy": 0.8527918781725888,
"type_macro_f1": 0.45105374473910137,
"span_exact": 0.6598984771573604,
"age_accuracy": 0.9796954314720813,
"loss": 1.8807837918323556,
"type_accuracy": 0.9211009174311927,
"type_macro_f1": 0.6878025559450562,
"span_exact": 0.5155963302752293,
"age_accuracy": 0.9486238532110092,
"age_f1": 0.0,
"joint_exact": 0.5685279187817259
"joint_exact": 0.44403669724770645
},
{
"epoch": 1,
"loss": 0.3179771784726564,
"type_accuracy": 0.9949238578680203,
"type_macro_f1": 0.9047619047619048,
"span_exact": 0.7918781725888325,
"loss": 0.19847860801096406,
"type_accuracy": 0.9743119266055046,
"type_macro_f1": 0.9018587930989455,
"span_exact": 0.7981651376146789,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.7868020304568528
"joint_exact": 0.7853211009174312
},
{
"epoch": 2,
"loss": 0.09052751894603979,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.8984771573604061,
"loss": 0.07505214461390836,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.8330275229357799,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.8984771573604061
"joint_exact": 0.8293577981651377
},
{
"epoch": 3,
"loss": 0.04577947580800937,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.8984771573604061,
"loss": 0.04440977191069969,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.8807339449541285,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.8984771573604061
"joint_exact": 0.8733944954128441
},
{
"epoch": 4,
"loss": 0.02791464072022889,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9086294416243654,
"loss": 0.028807769589840285,
"type_accuracy": 0.9889908256880734,
"type_macro_f1": 0.9096810438971369,
"span_exact": 0.8660550458715597,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9086294416243654
"joint_exact": 0.8605504587155963
},
{
"epoch": 5,
"loss": 0.018031590601412562,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9187817258883249,
"loss": 0.022258058590862728,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9155121981076735,
"span_exact": 0.8954128440366973,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9187817258883249
"joint_exact": 0.8880733944954129
},
{
"epoch": 6,
"loss": 0.01446260221151484,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9238578680203046,
"loss": 0.017832277844529697,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9155121981076735,
"span_exact": 0.9045871559633027,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9238578680203046
"joint_exact": 0.8972477064220183
},
{
"epoch": 7,
"loss": 0.010811449653628442,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9187817258883249,
"loss": 0.014963243124099362,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.8862385321100917,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9187817258883249
"joint_exact": 0.8788990825688073
},
{
"epoch": 8,
"loss": 0.00876794363632127,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9187817258883249,
"loss": 0.012998920554111813,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.9119266055045872,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9187817258883249
"joint_exact": 0.9045871559633027
},
{
"epoch": 9,
"loss": 0.007425106120052504,
"type_accuracy": 1.0,
"type_macro_f1": 1.0,
"span_exact": 0.9187817258883249,
"loss": 0.011474552309777529,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.9100917431192661,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9187817258883249
"joint_exact": 0.9027522935779817
},
{
"epoch": 10,
"loss": 0.010157171327384295,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.9027522935779817,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.8954128440366973
},
{
"epoch": 11,
"loss": 0.009997131634706275,
"type_accuracy": 0.9908256880733946,
"type_macro_f1": 0.9158000930443979,
"span_exact": 0.9045871559633027,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.8972477064220183
}
],
"test": {
"type_accuracy": 0.998062015503876,
"type_macro_f1": 0.9988596309351027,
"span_exact": 0.9244186046511628,
"type_accuracy": 0.9941699907947223,
"type_macro_f1": 0.9576679877290362,
"span_exact": 0.941699907947223,
"age_accuracy": 1.0,
"age_f1": 1.0,
"joint_exact": 0.9244186046511628
"joint_exact": 0.9389383246394599
},
"cv_mean": {
"type_accuracy": 0.992359410461513,
"type_macro_f1": 0.9481550900377457,
"span_exact": 0.9389851791472488,
"age_accuracy": 0.9996027805362463,
"age_f1": 0.9903614457831325,
"joint_exact": 0.9332879013203975
"type_accuracy": 0.9945265233277791,
"type_macro_f1": 0.9549157743958997,
"span_exact": 0.9167176505761322,
"age_accuracy": 0.9994757007005666,
"age_f1": 0.9892257817741743,
"joint_exact": 0.9125275899299806
},
"train_seconds": 156.0
"train_seconds": 589.2
}