wav2vec2-large-mms-1b-barwar
This model is a fine-tuned version of facebook/mms-1b-all on the nena_speech_1_0_test dataset. It achieves the following results on the evaluation set:
- Loss: 3.4355
- Wer: 1.0
- Cer: 0.3295
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: 32
- eval_batch_size: 8
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- num_epochs: 5
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
16.2134 | 0.15 | 25 | 17.6705 | 1.0 | 1.1562 |
6.1907 | 0.3 | 50 | 4.2262 | 0.9815 | 0.9986 |
3.6438 | 0.45 | 75 | 4.1625 | 1.0 | 0.7804 |
3.1627 | 0.6 | 100 | 4.1537 | 1.0 | 0.4727 |
2.2322 | 0.75 | 125 | 3.8028 | 1.0 | 0.4554 |
3.0705 | 0.9 | 150 | 3.3680 | 1.0 | 0.4352 |
2.9176 | 1.05 | 175 | 3.2934 | 1.0 | 0.4279 |
2.2255 | 1.2 | 200 | 3.6359 | 1.0 | 0.3926 |
2.4518 | 1.35 | 225 | 3.2249 | 1.0 | 0.3863 |
1.9254 | 1.5 | 250 | 3.7029 | 1.0 | 0.3875 |
2.7212 | 1.65 | 275 | 3.5201 | 1.0 | 0.3673 |
2.7976 | 1.8 | 300 | 3.3253 | 1.0 | 0.3986 |
2.0545 | 1.95 | 325 | 3.6138 | 1.0 | 0.3554 |
2.3335 | 2.1 | 350 | 3.5161 | 1.0 | 0.3554 |
2.0049 | 2.25 | 375 | 3.4727 | 1.0 | 0.3543 |
2.8896 | 2.4 | 400 | 3.2484 | 1.0 | 0.3535 |
1.9641 | 2.54 | 425 | 3.4330 | 1.0 | 0.3485 |
1.9649 | 2.69 | 450 | 3.8596 | 1.0 | 0.3444 |
2.0422 | 2.84 | 475 | 3.4291 | 1.0 | 0.3506 |
2.4093 | 2.99 | 500 | 3.3137 | 1.0 | 0.3434 |
1.8187 | 3.14 | 525 | 3.4423 | 1.0 | 0.3415 |
1.7495 | 3.29 | 550 | 3.5614 | 1.0 | 0.3431 |
2.0658 | 3.44 | 575 | 3.0324 | 1.0 | 0.3543 |
1.5128 | 3.59 | 600 | 3.6654 | 1.0 | 0.3452 |
1.7876 | 3.74 | 625 | 3.8747 | 1.0 | 0.3388 |
3.8652 | 3.89 | 650 | 2.9874 | 1.0 | 0.3387 |
2.8945 | 4.04 | 675 | 3.3015 | 1.0 | 0.3344 |
1.9763 | 4.19 | 700 | 3.1970 | 1.0 | 0.3389 |
2.0538 | 4.34 | 725 | 3.4811 | 1.0 | 0.3316 |
1.7723 | 4.49 | 750 | 3.6706 | 1.0 | 0.3305 |
2.0489 | 4.64 | 775 | 3.4281 | 1.0 | 0.3312 |
2.555 | 4.79 | 800 | 3.2610 | 1.0 | 0.3341 |
1.6591 | 4.94 | 825 | 3.4355 | 1.0 | 0.3295 |
Framework versions
- Transformers 4.34.1
- Pytorch 2.1.0+cu121
- Datasets 2.14.6
- Tokenizers 0.14.1
- Downloads last month
- 3
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 mnazari/wav2vec2-large-mms-1b-barwar
Base model
facebook/mms-1b-all
Finetuned
this model
Evaluation results
- Wer on nena_speech_1_0_testtest set self-reported1.000