wav2vec2-large-xls-r-300m-sw-10hr-v1

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

• Loss: 0.3209
• Wer: 0.2402
• Cer: 0.0623

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: 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: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
5.1124	0.9732	200	2.8417	1.0	1.0
1.3729	1.9465	400	0.5626	0.5369	0.1395
0.4189	2.9197	600	0.3993	0.4045	0.1012
0.2836	3.8929	800	0.3492	0.3428	0.0863
0.202	4.8662	1000	0.3426	0.3230	0.0812
0.1529	5.8394	1200	0.3318	0.2909	0.0739
0.1171	6.8127	1400	0.3149	0.2759	0.0710
0.0957	7.7859	1600	0.3239	0.2641	0.0681
0.0761	8.7591	1800	0.3383	0.2548	0.0658
0.0639	9.7324	2000	0.3204	0.2398	0.0624

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1