wav2vec2-xlsr-53-ft-cv-cy

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: 0.3657
• Wer: 0.3209

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: 4
• 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: 1000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.1004	200	3.0630	1.0
No log	0.2008	400	2.8947	1.0
4.9602	0.3012	600	1.3590	1.0133
4.9602	0.4016	800	0.9471	0.8022
1.1386	0.5020	1000	0.8021	0.7323
1.1386	0.6024	1200	0.7591	0.7313
1.1386	0.7028	1400	0.6862	0.6640
0.7718	0.8032	1600	0.6846	0.6641
0.7718	0.9036	1800	0.6149	0.6094
0.648	1.0040	2000	0.5827	0.6004
0.648	1.1044	2200	0.5644	0.5791
0.648	1.2048	2400	0.5603	0.6075
0.5109	1.3052	2600	0.5172	0.5455
0.5109	1.4056	2800	0.5177	0.5344
0.4791	1.5060	3000	0.4826	0.5282
0.4791	1.6064	3200	0.4892	0.5087
0.4791	1.7068	3400	0.4955	0.5177
0.456	1.8072	3600	0.4606	0.4787
0.456	1.9076	3800	0.4503	0.4542
0.4231	2.0080	4000	0.4367	0.4458
0.4231	2.1084	4200	0.4310	0.4299
0.4231	2.2088	4400	0.4280	0.4280
0.3026	2.3092	4600	0.4275	0.4354
0.3026	2.4096	4800	0.4091	0.4190
0.302	2.5100	5000	0.4177	0.4192
0.302	2.6104	5200	0.3975	0.4017
0.302	2.7108	5400	0.3830	0.3923
0.2939	2.8112	5600	0.3870	0.3908
0.2939	2.9116	5800	0.3951	0.3824
0.27	3.0120	6000	0.3864	0.3755
0.27	3.1124	6200	0.4179	0.3816
0.27	3.2129	6400	0.3908	0.3714
0.1997	3.3133	6600	0.3918	0.3745
0.1997	3.4137	6800	0.3852	0.3707
0.1937	3.5141	7000	0.3830	0.3645
0.1937	3.6145	7200	0.3788	0.3583
0.1937	3.7149	7400	0.3864	0.3570
0.1845	3.8153	7600	0.3664	0.3489
0.1845	3.9157	7800	0.3582	0.3461
0.1744	4.0161	8000	0.3701	0.3422
0.1744	4.1165	8200	0.3785	0.3368
0.1744	4.2169	8400	0.3786	0.3363
0.1325	4.3173	8600	0.3853	0.3354
0.1325	4.4177	8800	0.3799	0.3367
0.1225	4.5181	9000	0.3710	0.3294
0.1225	4.6185	9200	0.3734	0.3268
0.1225	4.7189	9400	0.3689	0.3243
0.1205	4.8193	9600	0.3646	0.3250
0.1205	4.9197	9800	0.3660	0.3215
0.118	5.0201	10000	0.3657	0.3209

Framework versions

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