<!-- 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. -->
wavlm2phone-jdrt-fw07-5N-holdout1
This model is a fine-tuned version of jonatasgrosman/exp_w2v2t_ja_wavlm_s729 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.6883
- Wer: 0.8427
- Cer: 0.7187
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: 1e-05
- train_batch_size: 64
- eval_batch_size: 32
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 1000
- num_epochs: 20
Training results
Training Loss | Epoch | Step | Validation Loss | Wer | Cer |
---|---|---|---|---|---|
10.1995 | 1.0 | 242 | 10.1039 | 1.0009 | 1.5886 |
9.181 | 2.0 | 484 | 8.9207 | 0.9632 | 0.7866 |
7.4004 | 3.0 | 726 | 6.9080 | 0.9076 | 0.8842 |
5.1218 | 4.0 | 968 | 4.6699 | 0.9072 | 0.9495 |
3.3049 | 5.0 | 1210 | 3.0563 | 0.9072 | 0.9495 |
2.3781 | 6.0 | 1452 | 2.2715 | 0.8813 | 0.8753 |
2.0545 | 7.0 | 1694 | 1.9466 | 0.8606 | 0.8113 |
1.8898 | 8.0 | 1936 | 1.8334 | 0.8595 | 0.8084 |
1.837 | 9.0 | 2178 | 1.7886 | 0.8569 | 0.7998 |
1.8117 | 10.0 | 2420 | 1.7616 | 0.8527 | 0.7848 |
1.7937 | 11.0 | 2662 | 1.7430 | 0.8488 | 0.7685 |
1.7741 | 12.0 | 2904 | 1.7288 | 0.8456 | 0.7541 |
1.767 | 13.0 | 3146 | 1.7178 | 0.8430 | 0.7430 |
1.7576 | 14.0 | 3388 | 1.7093 | 0.8421 | 0.7351 |
1.751 | 15.0 | 3630 | 1.7025 | 0.8421 | 0.7289 |
1.7553 | 16.0 | 3872 | 1.6973 | 0.8423 | 0.7248 |
1.7446 | 17.0 | 4114 | 1.6934 | 0.8419 | 0.7217 |
1.7388 | 18.0 | 4356 | 1.6906 | 0.8422 | 0.7197 |
1.7319 | 19.0 | 4598 | 1.6889 | 0.8425 | 0.7190 |
1.74 | 20.0 | 4840 | 1.6883 | 0.8427 | 0.7187 |
Framework versions
- Transformers 4.34.0.dev0
- Pytorch 2.0.1+cu118
- Datasets 2.14.5
- Tokenizers 0.13.3