Edit model card

hubert-large-ls960-ft-lg-CV-Fleurs_yogera_filtered-50hrs-v1

This model is a fine-tuned version of facebook/hubert-large-ls960-ft on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4719
  • Wer: 0.1529
  • Cer: 0.0370

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: 0.0003
  • train_batch_size: 8
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 16
  • 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
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
1.0507 1.0 1240 0.5309 0.5423 0.1256
0.4736 2.0 2480 0.4286 0.4675 0.1107
0.3919 3.0 3720 0.3794 0.4172 0.0960
0.3432 4.0 4960 0.3653 0.3843 0.0912
0.3091 5.0 6200 0.3767 0.3680 0.0835
0.2817 6.0 7440 0.3332 0.3569 0.0816
0.2604 7.0 8680 0.3358 0.3570 0.0782
0.243 8.0 9920 0.3278 0.3430 0.0793
0.2251 9.0 11160 0.3310 0.3238 0.0729
0.213 10.0 12400 0.3062 0.3068 0.0686
0.2008 11.0 13640 0.3290 0.3022 0.0670
0.1877 12.0 14880 0.3553 0.3010 0.0657
0.1775 13.0 16120 0.3821 0.2911 0.0651
0.1694 14.0 17360 0.3288 0.2759 0.0628
0.1609 15.0 18600 0.3198 0.2653 0.0598
0.1521 16.0 19840 0.3580 0.2610 0.0596
0.1463 17.0 21080 0.3339 0.2555 0.0582
0.1407 18.0 22320 0.3099 0.2521 0.0568
0.1318 19.0 23560 0.3401 0.2435 0.0560
0.1253 20.0 24800 0.3509 0.2458 0.0566
0.1212 21.0 26040 0.3210 0.2408 0.0552
0.1154 22.0 27280 0.3315 0.2324 0.0531
0.1122 23.0 28520 0.3424 0.2270 0.0527
0.1073 24.0 29760 0.3348 0.2206 0.0512
0.1025 25.0 31000 0.3673 0.2186 0.0516
0.1018 26.0 32240 0.3763 0.2112 0.0500
0.0968 27.0 33480 0.3833 0.2165 0.0514
0.0945 28.0 34720 0.3651 0.2143 0.0505
0.0911 29.0 35960 0.3487 0.2040 0.0482
0.0874 30.0 37200 0.3601 0.2135 0.0501
0.0864 31.0 38440 0.3620 0.2051 0.0490
0.0822 32.0 39680 0.3670 0.1977 0.0469
0.0811 33.0 40920 0.3617 0.2069 0.0482
0.0785 34.0 42160 0.3388 0.1993 0.0467
0.0767 35.0 43400 0.3688 0.1949 0.0461
0.0744 36.0 44640 0.3947 0.1970 0.0460
0.0721 37.0 45880 0.3893 0.1935 0.0460
0.0705 38.0 47120 0.3777 0.1909 0.0456
0.069 39.0 48360 0.3849 0.1883 0.0450
0.0671 40.0 49600 0.3988 0.1857 0.0445
0.0668 41.0 50840 0.4149 0.1840 0.0444
0.0645 42.0 52080 0.3914 0.1855 0.0444
0.0639 43.0 53320 0.3858 0.1852 0.0435
0.0594 44.0 54560 0.4093 0.1829 0.0440
0.0616 45.0 55800 0.3927 0.1806 0.0438
0.0588 46.0 57040 0.3734 0.1809 0.0436
0.0581 47.0 58280 0.3860 0.1830 0.0443
0.0568 48.0 59520 0.4135 0.1764 0.0429
0.0546 49.0 60760 0.4290 0.1783 0.0430
0.0529 50.0 62000 0.4386 0.1784 0.0431
0.0527 51.0 63240 0.4352 0.1811 0.0438
0.0521 52.0 64480 0.4121 0.1761 0.0425
0.0512 53.0 65720 0.3821 0.1743 0.0418
0.0511 54.0 66960 0.3834 0.1700 0.0415
0.0498 55.0 68200 0.3893 0.1714 0.0416
0.0464 56.0 69440 0.4116 0.1744 0.0418
0.046 57.0 70680 0.4081 0.1732 0.0419
0.0446 58.0 71920 0.4176 0.1712 0.0412
0.0446 59.0 73160 0.4000 0.1719 0.0416
0.0447 60.0 74400 0.4184 0.1750 0.0417
0.0436 61.0 75640 0.3897 0.1732 0.0416
0.0425 62.0 76880 0.4207 0.1687 0.0403
0.0406 63.0 78120 0.4006 0.1671 0.0403
0.0406 64.0 79360 0.3836 0.1698 0.0405
0.0411 65.0 80600 0.3942 0.1686 0.0406
0.0393 66.0 81840 0.3898 0.1653 0.0396
0.0392 67.0 83080 0.4115 0.1689 0.0405
0.0373 68.0 84320 0.4178 0.1663 0.0397
0.0365 69.0 85560 0.3999 0.1652 0.0397
0.0356 70.0 86800 0.4286 0.1663 0.0404
0.035 71.0 88040 0.4372 0.1621 0.0393
0.0339 72.0 89280 0.4113 0.1652 0.0394
0.0339 73.0 90520 0.4260 0.1641 0.0399
0.0343 74.0 91760 0.4216 0.1675 0.0399
0.033 75.0 93000 0.4403 0.1631 0.0394
0.0316 76.0 94240 0.4502 0.1626 0.0393
0.0323 77.0 95480 0.4378 0.1613 0.0390
0.0294 78.0 96720 0.4308 0.1622 0.0394
0.0303 79.0 97960 0.4377 0.1618 0.0390
0.0293 80.0 99200 0.4432 0.1575 0.0386
0.0291 81.0 100440 0.4232 0.1571 0.0386
0.0281 82.0 101680 0.4301 0.1602 0.0391
0.028 83.0 102920 0.4493 0.1605 0.0392
0.0276 84.0 104160 0.4431 0.1572 0.0384
0.0261 85.0 105400 0.4463 0.1554 0.0375
0.026 86.0 106640 0.4291 0.1562 0.0375
0.0263 87.0 107880 0.4337 0.1579 0.0379
0.0257 88.0 109120 0.4381 0.1546 0.0375
0.0254 89.0 110360 0.4724 0.1573 0.0379
0.0247 90.0 111600 0.4676 0.1564 0.0376
0.0243 91.0 112840 0.4500 0.1550 0.0373
0.0241 92.0 114080 0.4428 0.1573 0.0381
0.023 93.0 115320 0.4458 0.1550 0.0379
0.0229 94.0 116560 0.4553 0.1550 0.0378
0.0232 95.0 117800 0.4592 0.1550 0.0377
0.0219 96.0 119040 0.4752 0.1519 0.0370
0.022 97.0 120280 0.4770 0.1531 0.0374
0.0222 98.0 121520 0.4745 0.1532 0.0370
0.0214 99.0 122760 0.4612 0.1529 0.0370
0.021 100.0 124000 0.4719 0.1529 0.0370

Framework versions

  • Transformers 4.46.3
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.20.3
Downloads last month
2
Safetensors
Model size
315M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for asr-africa/hubert-large-ls960-ft-lg-CV-Fleurs_yogera_filtered-50hrs-v1

Finetuned
(25)
this model