metadata
license: apache-2.0
tags:
- generated_from_trainer
datasets:
- common_voice
model-index:
- name: wav2vec2-large-xls-r-300m-hi-wx1
results: []
wav2vec2-large-xls-r-300m-hi-wx1
This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice dataset. It achieves the following results on the evaluation set:
- Loss: 0.6966
- Wer: 0.3424
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: 0.00024
- train_batch_size: 16
- eval_batch_size: 8
- 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: 1500
- num_epochs: 50
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
11.0193 | 1.36 | 200 | 5.6992 | 1.0 |
3.9285 | 2.72 | 400 | 3.4430 | 1.0 |
3.293 | 4.08 | 600 | 2.7707 | 1.0 |
1.2714 | 5.44 | 800 | 0.8479 | 0.6015 |
0.6469 | 6.8 | 1000 | 0.7087 | 0.5412 |
0.4982 | 8.16 | 1200 | 0.6655 | 0.5116 |
0.4131 | 9.52 | 1400 | 0.6422 | 0.4845 |
0.3653 | 10.88 | 1600 | 0.6452 | 0.4700 |
0.3177 | 12.24 | 1800 | 0.7519 | 0.5236 |
0.2771 | 13.6 | 2000 | 0.6238 | 0.4444 |
0.2424 | 14.96 | 2200 | 0.6317 | 0.4458 |
0.2207 | 16.33 | 2400 | 0.6634 | 0.4144 |
0.1997 | 17.68 | 2600 | 0.6469 | 0.4110 |
0.1843 | 19.05 | 2800 | 0.6958 | 0.4162 |
0.1705 | 20.41 | 3000 | 0.6658 | 0.3992 |
0.1535 | 21.77 | 3200 | 0.6829 | 0.4136 |
0.1492 | 23.13 | 3400 | 0.6628 | 0.4018 |
0.1383 | 24.49 | 3600 | 0.6603 | 0.4020 |
0.1334 | 25.85 | 3800 | 0.7079 | 0.3914 |
0.1215 | 27.21 | 4000 | 0.7016 | 0.3904 |
0.1211 | 28.57 | 4200 | 0.7232 | 0.3963 |
0.1125 | 29.93 | 4400 | 0.7258 | 0.3879 |
0.1074 | 31.29 | 4600 | 0.7476 | 0.3900 |
0.0965 | 32.65 | 4800 | 0.7120 | 0.3734 |
0.0916 | 34.01 | 5000 | 0.6694 | 0.3730 |
0.0865 | 35.37 | 5200 | 0.7181 | 0.3680 |
0.0822 | 36.73 | 5400 | 0.6698 | 0.3554 |
0.0757 | 38.09 | 5600 | 0.7035 | 0.3627 |
0.0723 | 39.45 | 5800 | 0.6832 | 0.3575 |
0.0746 | 40.81 | 6000 | 0.6942 | 0.3508 |
0.0667 | 42.18 | 6200 | 0.7075 | 0.3523 |
0.0638 | 43.54 | 6400 | 0.7009 | 0.3473 |
0.0585 | 44.89 | 6600 | 0.6887 | 0.3443 |
0.0562 | 46.26 | 6800 | 0.6888 | 0.3454 |
0.0534 | 47.62 | 7000 | 0.7048 | 0.3438 |
0.0481 | 48.98 | 7200 | 0.6966 | 0.3424 |
Framework versions
- Transformers 4.16.2
- Pytorch 1.10.0+cu111
- Datasets 1.18.3
- Tokenizers 0.11.0