Hugging Face's logo

asr-malayalam
/
malayalam_combined_

TensorBoard
Safetensors
wav2vec2-bert
Generated from Trainer
Model card Files Metrics Community
malayalam_combined_ / README.md
akrishnan's picture
akrishnan
Training in progress, step 500
fcafedd verified
preview code
|
raw
history blame
4.32 kB
metadata
license: mit
base_model: facebook/w2v-bert-2.0
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: malayalam_combined_
    results: []

Visualize in Weights & Biases

malayalam_combined_

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4789
  • Wer: 0.4611

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: 5e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • 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: 50
  • num_epochs: 25
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
0.8401 0.2031 500 0.8498 0.7068
0.7367 0.4063 1000 0.7280 0.6183
0.6974 0.6094 1500 0.7055 0.6113
0.6493 0.8125 2000 0.6765 0.5989
0.5905 1.0156 2500 0.6521 0.5937
0.606 1.2188 3000 0.6192 0.5639
0.5601 1.4219 3500 0.6242 0.5526
0.5868 1.6250 4000 0.6118 0.5559
0.5792 1.8282 4500 0.5879 0.5523
0.554 2.0313 5000 0.5775 0.5501
0.505 2.2344 5500 0.5640 0.5466
0.5055 2.4375 6000 0.5668 0.5298
0.5228 2.6407 6500 0.5410 0.5178
0.5186 2.8438 7000 0.5785 0.5540
0.4811 3.0469 7500 0.5446 0.5408
0.4794 3.2501 8000 0.5333 0.5102
0.4952 3.4532 8500 0.5205 0.5135
0.4761 3.6563 9000 0.5218 0.5092
0.5079 3.8594 9500 0.5192 0.5166
0.4407 4.0626 10000 0.5207 0.5054
0.4711 4.2657 10500 0.5215 0.5086
0.4396 4.4688 11000 0.5289 0.5145
0.4667 4.6719 11500 0.5144 0.5015
0.4518 4.8751 12000 0.5222 0.5112
0.4211 5.0782 12500 0.5094 0.4897
0.43 5.2813 13000 0.5242 0.5011
0.4218 5.4845 13500 0.5132 0.4905
0.4279 5.6876 14000 0.5153 0.4883
0.4341 5.8907 14500 0.5321 0.4899
0.409 6.0938 15000 0.5079 0.4884
0.4111 6.2970 15500 0.5067 0.4844
0.3781 6.5001 16000 0.5091 0.4643
0.4274 6.7032 16500 0.4842 0.4831
0.4009 6.9064 17000 0.4791 0.4738
0.3895 7.1095 17500 0.4786 0.4691
0.3788 7.3126 18000 0.4845 0.4691
0.3909 7.5157 18500 0.4869 0.4612
0.3795 7.7189 19000 0.4729 0.4606
0.3874 7.9220 19500 0.4667 0.4655
0.3472 8.1251 20000 0.4718 0.4720
0.3634 8.3283 20500 0.4767 0.4616
0.3545 8.5314 21000 0.4821 0.4640
0.37 8.7345 21500 0.4789 0.4611

Framework versions

  • Transformers 4.43.0.dev0
  • Pytorch 1.14.0a0+44dac51
  • Datasets 2.16.1
  • Tokenizers 0.19.1