he-cantillation

This model is a fine-tuned version of ivrit-ai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 5.0244
  • Wer: 97.8059
  • Avg Precision Exact: 0.0463
  • Avg Recall Exact: 0.1014
  • Avg F1 Exact: 0.0598
  • Avg Precision Letter Shift: 0.0622
  • Avg Recall Letter Shift: 0.1383
  • Avg F1 Letter Shift: 0.0805
  • Avg Precision Word Level: 0.0777
  • Avg Recall Word Level: 0.1656
  • Avg F1 Word Level: 0.0970
  • Avg Precision Word Shift: 0.1542
  • Avg Recall Word Shift: 0.3497
  • Avg F1 Word Shift: 0.1988
  • Precision Median Exact: 0.0227
  • Recall Median Exact: 0.0625
  • F1 Median Exact: 0.0357
  • Precision Max Exact: 1.0
  • Recall Max Exact: 1.0
  • F1 Max Exact: 1.0
  • Precision Min Exact: 0.0
  • Recall Min Exact: 0.0
  • F1 Min Exact: 0.0
  • Precision Min Letter Shift: 0.0
  • Recall Min Letter Shift: 0.0
  • F1 Min Letter Shift: 0.0
  • Precision Min Word Level: 0.0
  • Recall Min Word Level: 0.0
  • F1 Min Word Level: 0.0
  • Precision Min Word Shift: 0.0
  • Recall Min Word Shift: 0.0
  • F1 Min Word Shift: 0.0

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-05
  • train_batch_size: 16
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • training_steps: 60000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Avg Precision Exact Avg Recall Exact Avg F1 Exact Avg Precision Letter Shift Avg Recall Letter Shift Avg F1 Letter Shift Avg Precision Word Level Avg Recall Word Level Avg F1 Word Level Avg Precision Word Shift Avg Recall Word Shift Avg F1 Word Shift Precision Median Exact Recall Median Exact F1 Median Exact Precision Max Exact Recall Max Exact F1 Max Exact Precision Min Exact Recall Min Exact F1 Min Exact Precision Min Letter Shift Recall Min Letter Shift F1 Min Letter Shift Precision Min Word Level Recall Min Word Level F1 Min Word Level Precision Min Word Shift Recall Min Word Shift F1 Min Word Shift
No log 0.0002 1 7.1581 109.4023 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0.0568 0.3754 2500 2.0970 96.9729 0.0613 0.0775 0.0671 0.0842 0.1078 0.0923 0.1037 0.1324 0.1135 0.2206 0.2995 0.2485 0.0345 0.05 0.04 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0314 0.7508 5000 2.3109 96.4492 0.0709 0.0953 0.0792 0.0932 0.1273 0.1045 0.1089 0.1502 0.1225 0.2270 0.3279 0.2605 0.0357 0.0588 0.0435 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.018 1.1261 7500 2.5849 97.4587 0.0458 0.0776 0.0548 0.0637 0.1096 0.0766 0.0775 0.1319 0.0926 0.1673 0.2959 0.2038 0.025 0.0455 0.0331 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0075 1.5015 10000 3.1851 97.9489 0.0397 0.0824 0.0507 0.0539 0.1151 0.0693 0.0664 0.1414 0.0847 0.1476 0.3210 0.1893 0.0204 0.05 0.0317 1.0 1.0 0.8571 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0153 1.8769 12500 2.7215 96.5921 0.0645 0.0950 0.0740 0.0868 0.1307 0.1001 0.1080 0.1624 0.1228 0.2293 0.3653 0.2688 0.0323 0.0625 0.0417 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0108 2.2523 15000 3.2672 97.2092 0.0544 0.1084 0.0685 0.0710 0.1448 0.0901 0.0848 0.1708 0.1064 0.1695 0.3629 0.2190 0.0270 0.0714 0.0392 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0077 2.6276 17500 3.2673 97.1625 0.0551 0.1014 0.0684 0.0749 0.1379 0.0922 0.0895 0.1631 0.1091 0.1845 0.3500 0.2292 0.0278 0.0625 0.0392 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0179 3.0030 20000 3.2682 97.2967 0.0516 0.0918 0.0634 0.0712 0.1274 0.0875 0.0869 0.1538 0.1061 0.1856 0.3430 0.2305 0.0278 0.0625 0.0385 1.0 0.8125 0.6667 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0066 3.3784 22500 3.5714 97.3726 0.0454 0.0955 0.0584 0.0621 0.1316 0.0796 0.0742 0.1559 0.0944 0.1542 0.3349 0.1987 0.0227 0.0556 0.0345 0.8571 1.0 0.8571 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0042 3.7538 25000 3.8114 97.9270 0.0404 0.0894 0.0525 0.0558 0.1247 0.0726 0.0677 0.1516 0.0883 0.1420 0.3347 0.1890 0.0222 0.0588 0.0345 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0014 4.1291 27500 4.0215 96.8212 0.0506 0.1030 0.0654 0.0667 0.1376 0.0863 0.0801 0.1627 0.1023 0.1651 0.3500 0.2142 0.0263 0.0667 0.0385 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0021 4.5045 30000 4.0509 98.0262 0.0395 0.0905 0.0523 0.0545 0.1243 0.0716 0.0666 0.1525 0.0873 0.1443 0.3387 0.1904 0.0217 0.0625 0.0345 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0028 4.8799 32500 3.9264 98.0685 0.0381 0.0844 0.0492 0.0513 0.1175 0.0667 0.0632 0.1427 0.0812 0.1319 0.3158 0.1729 0.0 0.0 0.0 0.7778 1.0 0.8750 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0027 5.2553 35000 4.5942 97.9532 0.0446 0.0991 0.0585 0.0593 0.1323 0.0770 0.0731 0.1586 0.0924 0.1465 0.3347 0.1893 0.0227 0.0667 0.0348 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0019 5.6306 37500 4.1899 97.8249 0.0437 0.1073 0.0581 0.0569 0.1405 0.0753 0.0679 0.1646 0.0884 0.1375 0.3415 0.1811 0.0204 0.0588 0.0331 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0011 6.0060 40000 4.5501 98.1939 0.0396 0.0917 0.0528 0.0555 0.1270 0.0728 0.0689 0.1550 0.0890 0.1460 0.3409 0.1913 0.0196 0.0588 0.0317 0.7778 1.0 0.8235 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0008 6.3814 42500 4.3012 97.9882 0.0421 0.0935 0.0552 0.0577 0.1284 0.0753 0.0710 0.1548 0.0909 0.1450 0.3337 0.1898 0.0217 0.0588 0.0339 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0007 6.7568 45000 4.2077 97.8409 0.0456 0.0924 0.0583 0.0623 0.1262 0.0792 0.0756 0.1512 0.0949 0.1578 0.3287 0.2014 0.0244 0.0625 0.0357 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0007 7.1321 47500 4.5387 97.6308 0.0510 0.1027 0.0652 0.0679 0.1377 0.0863 0.0815 0.1632 0.1026 0.1655 0.3474 0.2120 0.0263 0.0667 0.0385 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.001 7.5075 50000 4.3632 98.0670 0.0457 0.0830 0.0565 0.0635 0.1169 0.0785 0.0792 0.1419 0.0955 0.1735 0.3260 0.2133 0.0253 0.0556 0.0357 1.0 1.0 0.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0004 7.8829 52500 4.4452 97.6060 0.0445 0.0961 0.0580 0.0599 0.1288 0.0772 0.0723 0.1532 0.0921 0.1513 0.3325 0.1962 0.0233 0.0625 0.0351 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0001 8.2583 55000 4.5731 97.7607 0.0475 0.1018 0.0614 0.0635 0.1380 0.0821 0.0787 0.1659 0.0992 0.1603 0.3549 0.2071 0.025 0.0667 0.0377 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0001 8.6336 57500 4.9017 97.6148 0.0474 0.1006 0.0609 0.0625 0.1370 0.0811 0.0777 0.1637 0.0978 0.1550 0.3454 0.2009 0.025 0.0667 0.0377 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0
0.0 9.0090 60000 5.0244 97.8059 0.0463 0.1014 0.0598 0.0622 0.1383 0.0805 0.0777 0.1656 0.0970 0.1542 0.3497 0.1988 0.0227 0.0625 0.0357 1.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0

Framework versions

  • Transformers 4.49.0
  • Pytorch 2.7.0+cu126
  • Datasets 2.12.0
  • Tokenizers 0.20.1
Downloads last month
1
Safetensors
Model size
809M params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for cantillation/Teamim-IvritAI-large-v3-turbo-new_WeightDecay-0.005_Augmented_date-07-05-2025

Finetuned
(4)
this model