metadata

library_name: transformers
license: apache-2.0
base_model: facebook/wav2vec2-base-960h
tags:
  - generated_from_trainer
datasets:
  - common_voice_17_0
metrics:
  - wer
model-index:
  - name: Wav2Vec-Urdu-Test4
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice_17_0
          type: common_voice_17_0
          config: ur
          split: test
          args: ur
        metrics:
          - name: Wer
            type: wer
            value: 99.17818453492715

Wav2Vec-Urdu-Test4

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

Loss: 1.0093
Wer: 99.1782

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: 4
eval_batch_size: 4
seed: 42
gradient_accumulation_steps: 4
total_train_batch_size: 16
optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_steps: 300
num_epochs: 10
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.0	19	18.2298	105.9395
19.0476	2.0	38	5.0072	100.0374
6.1409	3.0	57	1.9519	98.3190
1.5743	4.0	76	1.0208	98.5805
1.5743	5.0	95	0.8330	99.3650
1.0023	6.0	114	0.7071	98.3937
0.8148	7.0	133	0.7718	95.5174
0.7533	8.0	152	0.6405	93.5002
0.7533	9.0	171	1.5573	95.2559
0.7768	9.48	180	1.0093	99.1782

Framework versions

Transformers 4.50.0
Pytorch 2.6.0+cu124
Datasets 3.4.1
Tokenizers 0.21.1