<!-- 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. -->
Nystrom-W2V2-100hrs-take-4-unfreeze-extractor-try-2
This model is a fine-tuned version of rohitp1/Nystrom-W2V2-100hrs-take-4-unfreeze-extractor on the None dataset. It achieves the following results on the evaluation set:
- Loss: 27.1915
- Wer: 0.0869
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: 0.0005
- train_batch_size: 4
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 64
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.2
- num_epochs: 100
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
23.1458 | 9.01 | 1000 | 28.9573 | 0.1039 |
32.7156 | 18.02 | 2000 | 25.6155 | 0.1218 |
43.506 | 27.03 | 3000 | 27.6332 | 0.1228 |
43.3608 | 36.04 | 4000 | 26.0539 | 0.1169 |
39.984 | 45.04 | 5000 | 25.9836 | 0.1137 |
35.1977 | 54.05 | 6000 | 26.2060 | 0.1077 |
30.1951 | 63.06 | 7000 | 27.0999 | 0.1033 |
25.7519 | 72.07 | 8000 | 27.8459 | 0.0964 |
22.1982 | 81.08 | 9000 | 27.9773 | 0.0908 |
20.0551 | 90.09 | 10000 | 27.4222 | 0.0884 |
19.4505 | 99.1 | 11000 | 27.1915 | 0.0869 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1
- Datasets 2.7.1
- Tokenizers 0.11.0