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-20-hours-Google-Fleurs-ALF-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: 0.2401
  • Wer: 0.0730
  • Cer: 0.0225

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: 8
  • seed: 42
  • gradient_accumulation_steps: 4
  • 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
1.4576 0.8193 400 0.3340 0.3406 0.0850
0.4894 1.6385 800 0.2932 0.2681 0.0677
0.4551 2.4578 1200 0.3200 0.2950 0.0793
0.451 3.2770 1600 0.3421 0.2868 0.0761
0.4472 4.0963 2000 0.3920 0.3382 0.0982
0.4692 4.9155 2400 0.4229 0.3618 0.0985
0.4472 5.7348 2800 0.3777 0.3163 0.0888
0.4095 6.5540 3200 0.3953 0.3080 0.0878
0.3774 7.3733 3600 0.3623 0.2871 0.0815
0.3605 8.1925 4000 0.3424 0.2962 0.0826
0.34 9.0118 4400 0.3163 0.2658 0.0754
0.3019 9.8310 4800 0.2908 0.2424 0.0694
0.2829 10.6503 5200 0.2661 0.2224 0.0602
0.2695 11.4695 5600 0.3034 0.2363 0.0675
0.2533 12.2888 6000 0.2752 0.2306 0.0663
0.2327 13.1080 6400 0.2491 0.2163 0.0622
0.2139 13.9273 6800 0.2472 0.2072 0.0601
0.1931 14.7465 7200 0.2159 0.1702 0.0482
0.1886 15.5658 7600 0.2375 0.1886 0.0546
0.1743 16.3850 8000 0.2391 0.1749 0.0500
0.1567 17.2043 8400 0.2412 0.1671 0.0480
0.1343 18.0236 8800 0.2241 0.1644 0.0473
0.125 18.8428 9200 0.1875 0.1478 0.0436
0.1125 19.6621 9600 0.2255 0.1496 0.0446
0.1012 20.4813 10000 0.2090 0.1452 0.0432
0.0944 21.3006 10400 0.2591 0.1480 0.0441
0.0888 22.1198 10800 0.2383 0.1444 0.0421
0.0832 22.9391 11200 0.2035 0.1346 0.0412
0.068 23.7583 11600 0.2047 0.1382 0.0407
0.0601 24.5776 12000 0.1975 0.1207 0.0366
0.0556 25.3968 12400 0.2059 0.1250 0.0360
0.0558 26.2161 12800 0.2440 0.1306 0.0389
0.0532 27.0353 13200 0.2188 0.1247 0.0376
0.0492 27.8546 13600 0.2063 0.1256 0.0372
0.0394 28.6738 14000 0.1864 0.1148 0.0344
0.0403 29.4931 14400 0.2006 0.1079 0.0329
0.0366 30.3123 14800 0.2068 0.1047 0.0313
0.0306 31.1316 15200 0.2064 0.1138 0.0347
0.0286 31.9508 15600 0.2331 0.1222 0.0369
0.023 32.7701 16000 0.2057 0.1052 0.0318
0.0242 33.5893 16400 0.2144 0.1086 0.0330
0.0238 34.4086 16800 0.1819 0.1025 0.0307
0.021 35.2279 17200 0.2386 0.1072 0.0323
0.0159 36.0471 17600 0.2019 0.0942 0.0292
0.0146 36.8664 18000 0.1899 0.0940 0.0287
0.0122 37.6856 18400 0.2013 0.0917 0.0285
0.0112 38.5049 18800 0.1965 0.0925 0.0275
0.0111 39.3241 19200 0.2132 0.0976 0.0298
0.0101 40.1434 19600 0.2201 0.0869 0.0269
0.0076 40.9626 20000 0.2189 0.0891 0.0275
0.0064 41.7819 20400 0.2042 0.0869 0.0261
0.0058 42.6011 20800 0.2124 0.0879 0.0269
0.0048 43.4204 21200 0.2107 0.0817 0.0247
0.0031 44.2396 21600 0.2024 0.0775 0.0236
0.0024 45.0589 22000 0.2303 0.0858 0.0264
0.0019 45.8781 22400 0.2250 0.0803 0.0243
0.0015 46.6974 22800 0.2250 0.0782 0.0238
0.0013 47.5166 23200 0.2257 0.0760 0.0232
0.0008 48.3359 23600 0.2394 0.0765 0.0232
0.0007 49.1551 24000 0.2373 0.0751 0.0230
0.0004 49.9744 24400 0.2401 0.0730 0.0225

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.17.0
  • Tokenizers 0.19.1
Downloads last month
0
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-20-hours-Google-Fleurs-ALF-dataset

Finetuned
(252)
this model