<!-- 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
This model is a fine-tuned version of rohitp1/Nystrom-W2V2-100hrs-take-3 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 27.1839
- Wer: 0.0915
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 |
---|---|---|---|---|
30.9159 | 9.01 | 1000 | 27.2422 | 0.1205 |
47.7101 | 18.02 | 2000 | 25.0538 | 0.1366 |
61.5118 | 27.03 | 3000 | 25.3063 | 0.1343 |
57.2966 | 36.04 | 4000 | 25.0429 | 0.1276 |
50.7161 | 45.04 | 5000 | 27.2507 | 0.1235 |
43.7605 | 54.05 | 6000 | 25.9086 | 0.1145 |
36.8698 | 63.06 | 7000 | 26.4890 | 0.1085 |
31.0921 | 72.07 | 8000 | 26.9372 | 0.1021 |
26.4249 | 81.08 | 9000 | 27.8031 | 0.0961 |
23.336 | 90.09 | 10000 | 27.2129 | 0.0928 |
22.2249 | 99.1 | 11000 | 27.1839 | 0.0915 |
Framework versions
- Transformers 4.24.0
- Pytorch 1.12.1
- Datasets 2.7.1
- Tokenizers 0.11.0