<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-base-stac-msa-local
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: 2.0671
- Wer: 0.7924
- Cer: 0.3289
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
Training: STAC + Tunisian MSA Test: CS DATA
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0001
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 2
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 30
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
|---|---|---|---|---|---|
| 1.4697 | 1.0 | 3773 | 1.8242 | 0.9395 | 0.5135 |
| 1.1644 | 2.0 | 7546 | 1.6306 | 0.8731 | 0.4446 |
| 0.9517 | 3.0 | 11319 | 1.4122 | 0.8587 | 0.4059 |
| 0.8563 | 4.0 | 15092 | 1.5409 | 0.8386 | 0.4034 |
| 0.7556 | 5.0 | 18865 | 1.4103 | 0.8247 | 0.3724 |
| 0.6841 | 6.0 | 22638 | 1.4608 | 0.8166 | 0.3735 |
| 0.5834 | 7.0 | 26411 | 1.5139 | 0.8113 | 0.3646 |
| 0.5607 | 8.0 | 30184 | 1.5303 | 0.8263 | 0.3797 |
| 0.5442 | 9.0 | 33957 | 1.3824 | 0.8198 | 0.3476 |
| 0.4584 | 10.0 | 37730 | 1.6412 | 0.8160 | 0.3576 |
| 0.4257 | 11.0 | 41503 | 1.5575 | 0.8003 | 0.3514 |
| 0.3631 | 12.0 | 45276 | 1.5776 | 0.8141 | 0.3454 |
| 0.3272 | 13.0 | 49049 | 1.5124 | 0.8127 | 0.3399 |
| 0.3348 | 14.0 | 52822 | 1.6733 | 0.7946 | 0.3398 |
| 0.3231 | 15.0 | 56595 | 1.5154 | 0.7987 | 0.3324 |
| 0.2556 | 16.0 | 60368 | 1.6161 | 0.7993 | 0.3402 |
| 0.238 | 17.0 | 64141 | 1.6126 | 0.7974 | 0.3329 |
| 0.2228 | 18.0 | 67914 | 1.7419 | 0.8014 | 0.3291 |
| 0.2129 | 19.0 | 71687 | 1.8394 | 0.8015 | 0.3374 |
| 0.1975 | 20.0 | 75460 | 1.9307 | 0.7928 | 0.3451 |
| 0.1981 | 21.0 | 79233 | 1.8700 | 0.8080 | 0.3375 |
| 0.1628 | 22.0 | 83006 | 1.9776 | 0.8061 | 0.3408 |
| 0.1462 | 23.0 | 86779 | 1.9090 | 0.8031 | 0.3306 |
| 0.1555 | 24.0 | 90552 | 1.9063 | 0.7878 | 0.3294 |
| 0.1515 | 25.0 | 94325 | 1.9632 | 0.7963 | 0.3278 |
| 0.1194 | 26.0 | 98098 | 1.9280 | 0.7991 | 0.3301 |
| 0.1219 | 27.0 | 101871 | 2.0248 | 0.7927 | 0.3329 |
| 0.1184 | 28.0 | 105644 | 2.0447 | 0.7903 | 0.3314 |
| 0.074 | 29.0 | 109417 | 2.0513 | 0.7910 | 0.3287 |
| 0.0836 | 30.0 | 113190 | 2.0671 | 0.7924 | 0.3289 |
Framework versions
- Transformers 4.17.0
- Pytorch 1.8.1+cu102
- Datasets 1.18.3
- Tokenizers 0.12.1
