mms-1b-all-lg-CV-v1

This model is a fine-tuned version of facebook/mms-1b-all on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

  • Loss: inf
  • Wer: 0.2420
  • Cer: 0.0619

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: 0.001
  • 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
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 100
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
0.2848 1.0 4442 inf 0.2783 0.0702
0.1923 2.0 8884 inf 0.2848 0.0713
0.1895 3.0 13326 inf 0.2710 0.0686
0.1873 4.0 17768 inf 0.2712 0.0675
0.1852 5.0 22210 inf 0.2710 0.0675
0.1828 6.0 26652 inf 0.2696 0.0695
0.1812 7.0 31094 inf 0.2756 0.0672
0.1804 8.0 35536 inf 0.2717 0.0675
0.1788 9.0 39978 inf 0.2592 0.0668
0.1778 10.0 44420 inf 0.2590 0.0666
0.1767 11.0 48862 inf 0.2619 0.0661
0.1755 12.0 53304 inf 0.2580 0.0657
0.1746 13.0 57746 inf 0.2545 0.0642
0.1738 14.0 62188 inf 0.2579 0.0655
0.1736 15.0 66630 inf 0.2565 0.0652
0.1727 16.0 71072 inf 0.2572 0.0650
0.172 17.0 75514 inf 0.2553 0.0644
0.1717 18.0 79956 inf 0.2520 0.0636
0.1706 19.0 84398 inf 0.2565 0.0645
0.1697 20.0 88840 inf 0.2516 0.0637
0.1696 21.0 93282 inf 0.2511 0.0632
0.1686 22.0 97724 inf 0.2687 0.0657
0.1688 23.0 102166 inf 0.2523 0.0639
0.1673 24.0 106608 inf 0.2529 0.0644
0.1671 25.0 111050 inf 0.2514 0.0637
0.1664 26.0 115492 inf 0.2508 0.0636
0.1666 27.0 119934 inf 0.2550 0.0637
0.1655 28.0 124376 inf 0.2515 0.0636
0.1649 29.0 128818 inf 0.2458 0.0628
0.1647 30.0 133260 inf 0.2517 0.0637
0.1641 31.0 137702 inf 0.2477 0.0628
0.1637 32.0 142144 inf 0.2461 0.0629
0.1629 33.0 146586 inf 0.2480 0.0635
0.1628 34.0 151028 inf 0.2498 0.0633
0.1625 35.0 155470 inf 0.2555 0.0638
0.1619 36.0 159912 inf 0.2498 0.0639
0.1625 37.0 164354 inf 0.2469 0.0628
0.1613 38.0 168796 inf 0.2443 0.0630
0.1618 39.0 173238 inf 0.2462 0.0635
0.1611 40.0 177680 inf 0.2419 0.0622
0.1604 41.0 182122 inf 0.2433 0.0627
0.1592 42.0 186564 inf 0.2443 0.0628
0.1598 43.0 191006 inf 0.2437 0.0629
0.1589 44.0 195448 inf 0.2444 0.0627
0.1589 45.0 199890 inf 0.2455 0.0628
0.158 46.0 204332 inf 0.2384 0.0619
0.1574 47.0 208774 inf 0.2455 0.0628
0.157 48.0 213216 inf 0.2444 0.0620
0.157 49.0 217658 inf 0.2428 0.0627
0.1557 50.0 222100 inf 0.2423 0.0622
0.1557 51.0 226542 inf 0.2420 0.0623
0.1558 52.0 230984 inf 0.2483 0.0626
0.1553 53.0 235426 inf 0.2406 0.0621
0.155 54.0 239868 inf 0.2445 0.0608
0.1545 55.0 244310 inf 0.2443 0.0617
0.1546 56.0 248752 inf 0.2419 0.0614
0.1537 57.0 253194 inf 0.2380 0.0612
0.1532 58.0 257636 inf 0.2420 0.0619

Framework versions

  • Transformers 4.47.0
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.21.0
Downloads last month
33
Safetensors
Model size
965M 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 sulaimank/mms-1b-all-lg-CV-v1

Finetuned
(214)
this model

Collection including sulaimank/mms-1b-all-lg-CV-v1

Evaluation results