This commit is contained in:
ci 2026-10-21 03:35:59 +00:00
parent 384c657525
commit bd41185968
6 changed files with 18252 additions and 26904 deletions

BIN
b.tar.gz

Binary file not shown.

300
canary_en.jsonl Normal file

File diff suppressed because one or more lines are too long

300
canary_zh.jsonl Normal file

File diff suppressed because one or more lines are too long

View File

@ -3,31 +3,6 @@ Some weights of ModernBertForSequenceClassification were not initialized from th
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
[jev] LoRA targets: ['Wqkv', 'Wo']
trainable params: 2,298,625 || all params: 309,829,634 || trainable%: 0.7419
[jev] train=37182 calib=5291
[jev] ep0 step100/2322 loss=0.2687
[jev] ep0 step200/2322 loss=0.3022
[jev] ep0 step300/2322 loss=0.3244
[jev] ep0 step400/2322 loss=0.2449
[jev] ep0 step500/2322 loss=0.2033
[jev] ep0 step600/2322 loss=0.2433
[jev] ep0 step700/2322 loss=0.2197
[jev] ep0 step800/2322 loss=0.1367
[jev] ep0 step900/2322 loss=0.1337
[jev] ep0 step1000/2322 loss=0.1400
[jev] ep0 step1100/2322 loss=0.1538
/usr/local/lib/python3.10/dist-packages/torch/_inductor/compile_fx.py:282: UserWarning: TensorFloat32 tensor cores for float32 matrix multiplication available but not enabled. Consider setting `torch.set_float32_matmul_precision('high')` for better performance.
warnings.warn(
[jev] epoch 0 calibration: {'auc': 0.9947, 'ece': 0.0182, 'n': 5291}
[jev] live-checkpoint 已写 /tmp/jd/model_live
[jev] ep1 step1200/2322 loss=0.2625
[jev] ep1 step1300/2322 loss=0.0523
[jev] ep1 step1400/2322 loss=0.0902
[jev] ep1 step1500/2322 loss=0.0737
[jev] ep1 step1600/2322 loss=0.1239
[jev] ep1 step1700/2322 loss=0.0809
[jev] ep1 step1800/2322 loss=0.2293
[jev] ep1 step1900/2322 loss=0.2075
[jev] ep1 step2000/2322 loss=0.1444
[jev] ep1 step2100/2322 loss=0.1464
[jev] ep1 step2200/2322 loss=0.1202
[jev] ep1 step2300/2322 loss=0.2464
[jev] train=27954 calib=5291
[jev] canaries: ['en', 'zh']
[jev] ep0 step100/1746 loss=0.2554

View File

@ -1,7 +1,8 @@
{
"tag": "v0.4",
"train": 37182,
"generic_neg_added": 804,
"upweight": 3,
"eval_label_fixed": 0
"tag": "v0.5",
"train": 27954,
"en_share": 0.42,
"evidence_aligned": 1120,
"canary_en": 300,
"canary_zh": 300
}

44514
train.jsonl

File diff suppressed because one or more lines are too long