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.

Visualize in Weights & Biases

w2v2-bert-Wolof-18-hour-ALFFA-dataset

This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.2583
  • Wer: 0.0840
  • Cer: 0.0249

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: 16
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 50

Training results

Training Loss Epoch Step Validation Loss Wer Cer
1.5722 0.8466 400 0.3788 0.3741 0.0935
0.4747 1.6931 800 0.3267 0.2948 0.0759
0.4418 2.5397 1200 0.3194 0.2907 0.0772
0.4311 3.3862 1600 0.3449 0.2997 0.0826
0.4416 4.2328 2000 0.4257 0.3593 0.1025
0.451 5.0794 2400 0.4580 0.3486 0.1008
0.4244 5.9259 2800 0.3760 0.2977 0.0835
0.3868 6.7725 3200 0.3756 0.3036 0.0875
0.3505 7.6190 3600 0.3905 0.2865 0.0812
0.3243 8.4656 4000 0.3123 0.2609 0.0729
0.3056 9.3122 4400 0.3160 0.2563 0.0725
0.2833 10.1587 4800 0.3261 0.2662 0.0753
0.2503 11.0053 5200 0.3089 0.2506 0.0715
0.2337 11.8519 5600 0.2994 0.2413 0.0714
0.2115 12.6984 6000 0.2453 0.1943 0.0553
0.1871 13.5450 6400 0.2801 0.2124 0.0615
0.1817 14.3915 6800 0.3009 0.2309 0.0685
0.1704 15.2381 7200 0.2517 0.1855 0.0544
0.1531 16.0847 7600 0.2857 0.1944 0.0589
0.1382 16.9312 8000 0.2925 0.1831 0.0550
0.1204 17.7778 8400 0.2939 0.2052 0.0608
0.108 18.6243 8800 0.2640 0.1814 0.0542
0.1029 19.4709 9200 0.2199 0.1473 0.0440
0.0877 20.3175 9600 0.2468 0.1594 0.0471
0.0799 21.1640 10000 0.2413 0.1641 0.0508
0.0772 22.0106 10400 0.2604 0.1586 0.0467
0.0666 22.8571 10800 0.2410 0.1553 0.0461
0.0584 23.7037 11200 0.2060 0.1298 0.0379
0.0476 24.5503 11600 0.2377 0.1387 0.0412
0.0503 25.3968 12000 0.2285 0.1398 0.0421
0.0452 26.2434 12400 0.2598 0.1610 0.0485
0.0411 27.0899 12800 0.2391 0.1407 0.0415
0.0329 27.9365 13200 0.2561 0.1391 0.0411
0.0307 28.7831 13600 0.2051 0.1251 0.0371
0.0273 29.6296 14000 0.2547 0.1331 0.0401
0.0217 30.4762 14400 0.2815 0.1406 0.0428
0.0232 31.3228 14800 0.2335 0.1205 0.0365
0.0204 32.1693 15200 0.2367 0.1231 0.0369
0.0182 33.0159 15600 0.2453 0.1299 0.0394
0.0145 33.8624 16000 0.2103 0.1106 0.0332
0.0127 34.7090 16400 0.2362 0.1290 0.0376
0.0134 35.5556 16800 0.2313 0.1127 0.0335
0.0138 36.4021 17200 0.2293 0.1071 0.0328
0.0093 37.2487 17600 0.2642 0.1153 0.0359
0.0106 38.0952 18000 0.2470 0.1132 0.0343
0.0091 38.9418 18400 0.2113 0.1015 0.0295
0.0063 39.7884 18800 0.2124 0.0959 0.0285
0.0052 40.6349 19200 0.2144 0.1014 0.0297
0.0056 41.4815 19600 0.2417 0.1016 0.0307
0.0034 42.3280 20000 0.2590 0.1020 0.0303
0.0022 43.1746 20400 0.2845 0.1010 0.0305
0.0023 44.0212 20800 0.2644 0.0950 0.0290
0.0018 44.8677 21200 0.2477 0.0903 0.0266
0.001 45.7143 21600 0.2554 0.0907 0.0272
0.0011 46.5608 22000 0.2583 0.0891 0.0268
0.0009 47.4074 22400 0.2409 0.0834 0.0249
0.0005 48.2540 22800 0.2613 0.0862 0.0257
0.0005 49.1005 23200 0.2629 0.0864 0.0255
0.0005 49.9471 23600 0.2583 0.0840 0.0249

Framework versions

  • Transformers 4.44.1
  • Pytorch 2.1.0+cu118
  • Datasets 2.17.0
  • Tokenizers 0.19.1
Downloads last month
12
Safetensors
Model size
606M params
Tensor type
F32
·
Inference Providers NEW
This model is not currently available via any of the supported Inference Providers.

Model tree for asr-africa/w2v2-bert-Wolof-18-hour-ALFFA-dataset

Finetuned
(252)
this model