metadata
license: apache-2.0
base_model: facebook/wav2vec2-xls-r-300m
tags:
- generated_from_trainer
metrics:
- wer
model-index:
- name: wav2vec2-xls-r-300m-bengali-macro
results: []
wav2vec2-xls-r-300m-bengali-macro
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: 2.3787
- Wer: 0.88
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: 0.0003
- train_batch_size: 16
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 1
Training results
| Training Loss | Epoch | Step | Validation Loss | Wer |
|---|---|---|---|---|
| 4.686 | 0.02 | 500 | 2.9368 | 0.9254 |
| 1.465 | 0.03 | 1000 | 1.6714 | 0.88 |
| 1.2139 | 0.05 | 1500 | 1.6254 | 0.8292 |
| 1.1463 | 0.07 | 2000 | 1.5170 | 0.8292 |
| 1.12 | 0.08 | 2500 | 1.4973 | 0.7966 |
| 1.0766 | 0.1 | 3000 | 1.5682 | 0.8129 |
| 1.0547 | 0.12 | 3500 | 1.3838 | 0.7458 |
| 1.0163 | 0.13 | 4000 | 1.6073 | 0.8685 |
| 1.0149 | 0.15 | 4500 | 1.3993 | 0.7247 |
| 1.0125 | 0.17 | 5000 | 1.4888 | 0.7749 |
| 0.9882 | 0.18 | 5500 | 1.3766 | 0.7444 |
| 0.9736 | 0.2 | 6000 | 1.5816 | 0.8027 |
| 0.9737 | 0.22 | 6500 | 1.5761 | 0.7783 |
| 0.9445 | 0.23 | 7000 | 1.3593 | 0.7505 |
| 0.9335 | 0.25 | 7500 | 1.3453 | 0.7247 |
| 0.931 | 0.27 | 8000 | 1.4024 | 0.7397 |
| 0.9389 | 0.28 | 8500 | 1.5973 | 0.8508 |
| 0.9152 | 0.3 | 9000 | 1.4021 | 0.7193 |
| 0.9042 | 0.32 | 9500 | 1.3642 | 0.7620 |
| 0.8962 | 0.33 | 10000 | 1.4298 | 0.7383 |
| 0.8767 | 0.35 | 10500 | 1.4478 | 0.7580 |
| 0.8853 | 0.37 | 11000 | 1.3255 | 0.7302 |
| 0.8739 | 0.38 | 11500 | 1.3791 | 0.7431 |
| 0.8597 | 0.4 | 12000 | 1.5847 | 0.8325 |
| 0.8815 | 0.42 | 12500 | 1.6785 | 0.8163 |
| 0.8736 | 0.43 | 13000 | 1.6222 | 0.7871 |
| 0.8643 | 0.45 | 13500 | 1.8635 | 0.8502 |
| 0.84 | 0.46 | 14000 | 1.4343 | 0.7803 |
| 0.8323 | 0.48 | 14500 | 1.7500 | 0.8427 |
| 0.8223 | 0.5 | 15000 | 1.6916 | 0.8278 |
| 0.827 | 0.51 | 15500 | 2.6214 | 0.9085 |
| 0.8149 | 0.53 | 16000 | 1.6750 | 0.8169 |
| 0.8149 | 0.55 | 16500 | 1.7646 | 0.8142 |
| 0.8032 | 0.56 | 17000 | 2.1347 | 0.8617 |
| 0.8005 | 0.58 | 17500 | 1.7216 | 0.8122 |
| 0.7956 | 0.6 | 18000 | 2.3053 | 0.8936 |
| 0.7888 | 0.61 | 18500 | 1.7773 | 0.8359 |
| 0.7919 | 0.63 | 19000 | 2.2394 | 0.8597 |
| 0.7888 | 0.65 | 19500 | 1.5470 | 0.7403 |
| 0.7721 | 0.66 | 20000 | 1.6034 | 0.7593 |
| 0.7603 | 0.68 | 20500 | 1.6808 | 0.7803 |
| 0.751 | 0.7 | 21000 | 1.7942 | 0.8217 |
| 0.7555 | 0.71 | 21500 | 1.9897 | 0.8441 |
| 0.7583 | 0.73 | 22000 | 2.3329 | 0.8576 |
| 0.7346 | 0.75 | 22500 | 2.2255 | 0.8515 |
| 0.754 | 0.76 | 23000 | 2.2606 | 0.8861 |
| 0.7309 | 0.78 | 23500 | 2.0292 | 0.8529 |
| 0.7351 | 0.8 | 24000 | 2.4471 | 0.8942 |
| 0.7456 | 0.81 | 24500 | 2.1406 | 0.8224 |
| 0.7229 | 0.83 | 25000 | 2.4474 | 0.8888 |
| 0.7253 | 0.85 | 25500 | 2.0324 | 0.8441 |
| 0.7109 | 0.86 | 26000 | 2.2594 | 0.8671 |
| 0.7316 | 0.88 | 26500 | 2.3887 | 0.8827 |
| 0.716 | 0.9 | 27000 | 2.4739 | 0.8915 |
| 0.7264 | 0.91 | 27500 | 2.4291 | 0.8922 |
| 0.701 | 0.93 | 28000 | 2.3306 | 0.8936 |
| 0.7025 | 0.95 | 28500 | 2.3172 | 0.8834 |
| 0.6963 | 0.96 | 29000 | 2.4020 | 0.8841 |
| 0.6952 | 0.98 | 29500 | 2.4324 | 0.8895 |
| 0.6985 | 1.0 | 30000 | 2.3787 | 0.88 |
Framework versions
- Transformers 4.33.0
- Pytorch 2.0.0
- Datasets 2.14.5
- Tokenizers 0.13.3