You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

w2v-bert-2.0-real-250-synth-250-hausa-v0.0

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.6923
  • Wer: 0.1308
  • Cer: 0.0385

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: 9e-05
  • train_batch_size: 64
  • eval_batch_size: 32
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 128
  • optimizer: Use 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_ratio: 0.025
  • num_epochs: 100.0
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
0.4005 1.0 4324 0.3025 0.2770 0.0704
0.1588 2.0 8648 0.3274 0.3038 0.0763
0.1588 3.0 12972 0.2865 0.2555 0.0672
0.1494 4.0 17296 0.2799 0.2452 0.0645
0.142 5.0 21620 0.2884 0.2548 0.0675
0.1359 6.0 25944 0.2767 0.2436 0.0647
0.1301 7.0 30268 0.2626 0.2446 0.0653
0.1232 8.0 34592 0.2529 0.2322 0.0608
0.12 9.0 38916 0.2600 0.2351 0.0620
0.1158 10.0 43240 0.2742 0.2431 0.0644
0.1117 11.0 47564 0.2565 0.2297 0.0603
0.1078 12.0 51888 0.2568 0.2257 0.0595
0.1038 13.0 56212 0.2483 0.2285 0.0595
0.0993 14.0 60536 0.2427 0.2217 0.0577
0.0974 15.0 64860 0.2494 0.2229 0.0588
0.0936 16.0 69184 0.2465 0.2216 0.0585
0.0884 17.0 73508 0.2413 0.2139 0.0563
0.0837 18.0 77832 0.2489 0.2154 0.0566
0.0801 19.0 82156 0.2489 0.2124 0.0562
0.0753 20.0 86480 0.2446 0.2077 0.0550
0.071 21.0 90804 0.2480 0.2038 0.0542
0.067 22.0 95128 0.2475 0.2057 0.0548
0.0631 23.0 99452 0.2518 0.1970 0.0528
0.0577 24.0 103776 0.2663 0.1917 0.0510
0.0544 25.0 108100 0.2536 0.1919 0.0516
0.0499 26.0 112424 0.2642 0.1895 0.0513
0.0458 27.0 116748 0.2671 0.1799 0.0487
0.0425 28.0 121072 0.2808 0.1797 0.0491
0.0396 29.0 125396 0.2732 0.1853 0.0508
0.0375 30.0 129720 0.2822 0.1776 0.0485
0.0341 31.0 134044 0.3013 0.1720 0.0474
0.032 32.0 138368 0.2974 0.1753 0.0485
0.0289 33.0 142692 0.2989 0.1730 0.0481
0.0275 34.0 147016 0.3099 0.1709 0.0476
0.0252 35.0 151340 0.3063 0.1680 0.0471
0.0232 36.0 155664 0.3232 0.1684 0.0474
0.0218 37.0 159988 0.3440 0.1618 0.0455
0.0204 38.0 164312 0.3251 0.1637 0.0462
0.0187 39.0 168636 0.3395 0.1593 0.0449
0.0174 40.0 172960 0.3553 0.1595 0.0452
0.0163 41.0 177284 0.3417 0.1582 0.0448
0.0155 42.0 181608 0.3699 0.1552 0.0442
0.0144 43.0 185932 0.3698 0.1560 0.0443
0.0139 44.0 190256 0.3543 0.1574 0.0447
0.0125 45.0 194580 0.3713 0.1567 0.0448
0.0122 46.0 198904 0.3797 0.1515 0.0432
0.0111 47.0 203228 0.3836 0.1520 0.0436
0.0105 48.0 207552 0.3989 0.1486 0.0425
0.0097 49.0 211876 0.4116 0.1490 0.0427
0.0091 50.0 216200 0.3927 0.1479 0.0425
0.0088 51.0 220524 0.3921 0.1464 0.0422
0.0083 52.0 224848 0.4105 0.1518 0.0439
0.0077 53.0 229172 0.4117 0.1488 0.0427
0.0072 54.0 233496 0.4022 0.1517 0.0436
0.0068 55.0 237820 0.4175 0.1433 0.0415
0.0064 56.0 242144 0.4209 0.1459 0.0420
0.0062 57.0 246468 0.4381 0.1440 0.0416
0.0056 58.0 250792 0.4322 0.1434 0.0414
0.0054 59.0 255116 0.4287 0.1453 0.0421
0.0051 60.0 259440 0.4335 0.1433 0.0416
0.0048 61.0 263764 0.4541 0.1433 0.0415
0.0046 62.0 268088 0.4574 0.1427 0.0413
0.0042 63.0 272412 0.4478 0.1444 0.0418
0.004 64.0 276736 0.4769 0.1396 0.0407
0.0038 65.0 281060 0.4573 0.1444 0.0420
0.0036 66.0 285384 0.4657 0.1401 0.0408
0.0033 67.0 289708 0.4791 0.1422 0.0415
0.0033 68.0 294032 0.4793 0.1399 0.0407
0.003 69.0 298356 0.4883 0.1384 0.0404
0.0027 70.0 302680 0.4866 0.1386 0.0404
0.0026 71.0 307004 0.5008 0.1374 0.0402
0.0024 72.0 311328 0.5000 0.1398 0.0409
0.0023 73.0 315652 0.5073 0.1388 0.0407
0.002 74.0 319976 0.5455 0.1379 0.0401
0.002 75.0 324300 0.5289 0.1374 0.0401
0.0018 76.0 328624 0.5519 0.1347 0.0394
0.0017 77.0 332948 0.5298 0.1352 0.0394
0.0015 78.0 337272 0.5382 0.1363 0.0400
0.0015 79.0 341596 0.5173 0.1391 0.0405
0.0013 80.0 345920 0.5414 0.1349 0.0395
0.0012 81.0 350244 0.5724 0.1352 0.0396
0.0011 82.0 354568 0.5710 0.1353 0.0396
0.0009 83.0 358892 0.6018 0.1328 0.0388
0.0009 84.0 363216 0.5961 0.1325 0.0389
0.0008 85.0 367540 0.6118 0.1328 0.0389
0.0007 86.0 371864 0.5842 0.1321 0.0386
0.0007 87.0 376188 0.6197 0.1315 0.0385
0.0006 88.0 380512 0.6307 0.1322 0.0388
0.0005 89.0 384836 0.6503 0.1322 0.0386
0.0004 90.0 389160 0.6500 0.1325 0.0389
0.0003 91.0 393484 0.6967 0.1312 0.0385
0.0003 92.0 397808 0.6835 0.1315 0.0387
0.0002 93.0 402132 0.6923 0.1308 0.0385

Framework versions

  • Transformers 4.51.3
  • Pytorch 2.6.0+cu124
  • Datasets 3.5.0
  • Tokenizers 0.21.1
Downloads last month
10
Safetensors
Model size
606M 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 asr-africa/w2v-bert-2.0-real-250-synth-250-hausa-v0.0

Finetuned
(300)
this model