wav2vec2-large-xlsr-53-breton

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9840
• Wer: 0.5852
• Cer: 0.2130

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: 6e-05
• 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_ratio: 0.08
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
11.8947	2.56	250	3.4769	1.0	0.9862
3.1668	5.13	500	3.0459	1.0	0.9862
2.6491	7.69	750	1.6416	0.9319	0.4441
1.4107	10.26	1000	1.1000	0.7751	0.2852
0.9989	12.82	1250	0.9827	0.7092	0.2578
0.8238	15.38	1500	0.9543	0.6864	0.2476
0.7193	17.95	1750	0.9241	0.6547	0.2371
0.6377	20.51	2000	0.9296	0.6452	0.2352
0.5865	23.08	2250	0.9287	0.6320	0.2301
0.541	25.64	2500	0.9359	0.6205	0.2231
0.4988	28.21	2750	0.9850	0.6149	0.2244
0.4691	30.77	3000	0.9566	0.6065	0.2192
0.4568	33.33	3250	0.9653	0.6019	0.2175
0.4485	35.9	3500	0.9760	0.5949	0.2175
0.4219	38.46	3750	0.9824	0.5926	0.2177
0.397	41.03	4000	0.9669	0.5885	0.2138
0.3912	43.59	4250	0.9857	0.5908	0.2145
0.3764	46.15	4500	0.9937	0.5886	0.2145
0.3742	48.72	4750	0.9840	0.5852	0.2130

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0