wav2vec2-xlsr-53-ft-btb-cv-cy-cand

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: inf
• Wer: 0.3598

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: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 800
• training_steps: 8000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.6973	0.0714	500	inf	1.0
1.448	0.1428	1000	inf	0.7574
1.053	0.2142	1500	inf	0.6584
0.9304	0.2856	2000	inf	0.5963
0.8755	0.3569	2500	inf	0.5946
0.8238	0.4283	3000	inf	0.5392
0.7819	0.4997	3500	inf	0.4967
0.729	0.5711	4000	inf	0.4834
0.6923	0.6425	4500	inf	0.4564
0.7052	0.7139	5000	inf	0.4346
0.6675	0.7853	5500	inf	0.4163
0.6217	0.8567	6000	inf	0.3962
0.5954	0.9280	6500	inf	0.3883
0.5687	0.9994	7000	inf	0.3746
0.477	1.0708	7500	inf	0.3647
0.4804	1.1422	8000	inf	0.3598

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1