wav2vec2-large-xls-r-300m-tsovatush-demo-colab-bryn-hauk-11-10

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.0335
• Wer: 0.4883
• Cer: 0.1275

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: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.9765	6.8966	400	1.4359	0.9369	0.2946
1.1792	13.7931	800	0.8249	0.5694	0.1605
0.6068	20.6897	1200	0.8757	0.5261	0.1442
0.3661	27.5862	1600	0.9834	0.5171	0.1372
0.2159	34.4828	2000	0.9937	0.5099	0.1385
0.1284	41.3793	2400	1.0792	0.5225	0.1376
0.0839	48.2759	2800	1.0335	0.4883	0.1275

Framework versions

• Transformers 4.44.2
• Pytorch 2.5.0+cu121
• Datasets 1.18.3
• Tokenizers 0.19.1