Edit model card

wav2vec2-xls-r-300m-CV_Fleurs_AMMI_ALFFA-sw-400hrs-v1

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4397
  • Wer: 0.1390
  • Cer: 0.0457

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
0.9438 1.0 16876 0.5068 0.2899 0.0911
0.5501 2.0 33752 0.3912 0.2672 0.0823
0.4835 3.0 50628 0.3540 0.2436 0.0787
0.4402 4.0 67504 0.3511 0.2350 0.0727
0.4052 5.0 84380 0.3013 0.2234 0.0700
0.3786 6.0 101256 0.3334 0.2205 0.0725
0.3567 7.0 118132 0.3268 0.2080 0.0653
0.3347 8.0 135008 0.3131 0.2018 0.0646
0.3184 9.0 151884 0.2761 0.1928 0.0623
0.3035 10.0 168760 0.2957 0.1899 0.0594
0.288 11.0 185636 0.2986 0.1969 0.0677
0.2741 12.0 202512 0.2925 0.1833 0.0576
0.2636 13.0 219388 0.3275 0.1812 0.0593
0.2568 14.0 236264 0.2794 0.1791 0.0568
0.2413 15.0 253140 0.2805 0.1828 0.0594
0.2311 16.0 270016 0.3014 0.1716 0.0576
0.2218 17.0 286892 0.2842 0.1718 0.0556
0.2136 18.0 303768 0.2858 0.1692 0.0553
0.2036 19.0 320644 0.2833 0.1675 0.0536
0.1967 20.0 337520 0.2876 0.1628 0.0533
0.1878 21.0 354396 0.3176 0.1597 0.0519
0.1811 22.0 371272 0.2850 0.1643 0.0545
0.1739 23.0 388148 0.2613 0.1633 0.0530
0.1681 24.0 405024 0.2933 0.1571 0.0511
0.1623 25.0 421900 0.2958 0.1581 0.0509
0.157 26.0 438776 0.3099 0.1564 0.0507
0.1511 27.0 455652 0.3269 0.1572 0.0517
0.1465 28.0 472528 0.2733 0.1546 0.0501
0.1412 29.0 489404 0.2930 0.1549 0.0501
0.1371 30.0 506280 0.2686 0.1543 0.0501
0.1325 31.0 523156 0.3153 0.1531 0.0494
0.1283 32.0 540032 0.2964 0.1517 0.0493
0.1245 33.0 556908 0.3827 0.1507 0.0492
0.1215 34.0 573784 0.3416 0.1503 0.0493
0.1174 35.0 590660 0.3232 0.1499 0.0480
0.1149 36.0 607536 0.3119 0.1511 0.0500
0.1115 37.0 624412 0.3400 0.1506 0.0485
0.1076 38.0 641288 0.3502 0.1500 0.0493
0.1051 39.0 658164 0.3428 0.1479 0.0487
0.1022 40.0 675040 0.3596 0.1451 0.0484
0.1002 41.0 691916 0.3349 0.1492 0.0494
0.0978 42.0 708792 0.3514 0.1415 0.0468
0.0951 43.0 725668 0.3689 0.1447 0.0475
0.0925 44.0 742544 0.3720 0.1427 0.0472
0.0907 45.0 759420 0.3896 0.1448 0.0481
0.0882 46.0 776296 0.3680 0.1454 0.0486
0.0865 47.0 793172 0.3689 0.1445 0.0476
0.0849 48.0 810048 0.3505 0.1400 0.0460
0.0826 49.0 826924 0.3827 0.1392 0.0467
0.0804 50.0 843800 0.3828 0.1460 0.0481
0.0789 51.0 860676 0.4070 0.1419 0.0474
0.0774 52.0 877552 0.4041 0.1426 0.0473
0.075 53.0 894428 0.4212 0.1418 0.0474
0.073 54.0 911304 0.4072 0.1400 0.0468
0.0718 55.0 928180 0.4032 0.1388 0.0460
0.0694 56.0 945056 0.4591 0.1432 0.0474
0.0685 57.0 961932 0.4426 0.1384 0.0460
0.0671 58.0 978808 0.4397 0.1390 0.0457

Framework versions

  • Transformers 4.46.1
  • Pytorch 2.1.0+cu118
  • Datasets 3.1.0
  • Tokenizers 0.20.1
Downloads last month
60
Safetensors
Model size
316M 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/wav2vec2-xls-r-300m-CV_Fleurs_AMMI_ALFFA-sw-400hrs-v1

Finetuned
(433)
this model