xlsr-english

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the librispeech_asr dataset. It achieves the following results on the evaluation set:

• Loss: 0.3098
• Wer: 0.1451

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.2453	2.37	400	0.5789	0.4447
0.3736	4.73	800	0.3737	0.2850
0.1712	7.1	1200	0.3038	0.2136
0.117	9.47	1600	0.3016	0.2072
0.0897	11.83	2000	0.3158	0.1920
0.074	14.2	2400	0.3137	0.1831
0.0595	16.57	2800	0.2967	0.1745
0.0493	18.93	3200	0.3192	0.1670
0.0413	21.3	3600	0.3176	0.1644
0.0322	23.67	4000	0.3079	0.1598
0.0296	26.04	4400	0.2978	0.1511
0.0235	28.4	4800	0.3098	0.1451

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.0+cu113
• Datasets 1.18.3
• Tokenizers 0.10.3