wav2vec2-large-xls-r-300m-tsovatush-demo-colab-bryn-hauk-8-6

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: 1.4337
• Wer: 0.5188

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
5.3164	4.1667	400	3.2339	None
2.1905	8.3333	800	1.1741	0.6596
1.2308	12.5	1200	1.1192	0.6275
0.8868	16.6667	1600	1.1354	0.5776
0.6593	20.8333	2000	1.2042	0.5831
0.466	25.0	2400	1.3671	0.5654
0.3319	29.1667	2800	1.3642	0.5388
0.2511	33.3333	3200	1.3517	0.5621
0.1895	37.5	3600	1.3898	0.5366
0.1413	41.6667	4000	1.4396	0.5277
0.1086	45.8333	4400	1.4229	0.5344
0.0897	50.0	4800	1.4337	0.5188

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 1.18.3
• Tokenizers 0.19.1