wav2vec2-xls-r-300m-Fleurs_AMMI_AFRIVOICE_LRSC-ln-50hrs-v2

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: 0.3618
• Wer: 0.1994
• Cer: 0.0632

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 2
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.4765	1.0	2847	2.7985	1.0	1.0
1.8884	2.0	5694	0.7557	0.5439	0.1510
0.9348	3.0	8541	0.5497	0.3536	0.1072
0.7562	4.0	11388	0.4460	0.3048	0.0929
0.6678	5.0	14235	0.4077	0.2995	0.0907
0.6169	6.0	17082	0.3893	0.2918	0.0880
0.58	7.0	19929	0.3855	0.2834	0.0861
0.5497	8.0	22776	0.3514	0.2613	0.0806
0.525	9.0	25623	0.3421	0.2539	0.0795
0.5049	10.0	28470	0.3359	0.2540	0.0786
0.4862	11.0	31317	0.3337	0.2534	0.0783
0.4715	12.0	34164	0.3294	0.2362	0.0743
0.4569	13.0	37011	0.3087	0.2406	0.0731
0.4463	14.0	39858	0.3288	0.2265	0.0712
0.4329	15.0	42705	0.3379	0.2310	0.0724
0.4245	16.0	45552	0.3287	0.2229	0.0700
0.4092	17.0	48399	0.3202	0.2169	0.0684
0.4016	18.0	51246	0.3114	0.2153	0.0682
0.3903	19.0	54093	0.3204	0.2128	0.0671
0.3831	20.0	56940	0.3112	0.2122	0.0670
0.3728	21.0	59787	0.3169	0.2112	0.0667
0.3681	22.0	62634	0.3080	0.2089	0.0662
0.3606	23.0	65481	0.3238	0.2097	0.0668
0.3535	24.0	68328	0.3144	0.2084	0.0658
0.3432	25.0	71175	0.3192	0.2055	0.0655
0.339	26.0	74022	0.3123	0.2070	0.0661
0.3349	27.0	76869	0.3244	0.2043	0.0651
0.3217	28.0	79716	0.3109	0.2082	0.0676
0.313	29.0	82563	0.3263	0.2066	0.0655
0.3079	30.0	85410	0.3456	0.1997	0.0640
0.3024	31.0	88257	0.3035	0.2087	0.0672
0.2966	32.0	91104	0.3459	0.1969	0.0634
0.289	33.0	93951	0.3178	0.2005	0.0646
0.2828	34.0	96798	0.3217	0.2040	0.0657
0.2794	35.0	99645	0.3214	0.2085	0.0671
0.2746	36.0	102492	0.3556	0.1988	0.0633
0.2701	37.0	105339	0.3509	0.1984	0.0638
0.2663	38.0	108186	0.3671	0.2004	0.0640
0.2663	39.0	111033	0.3396	0.1980	0.0637
0.2572	40.0	113880	0.3707	0.2008	0.0643
0.2545	41.0	116727	0.3527	0.1970	0.0628
0.2526	42.0	119574	0.3618	0.1994	0.0632

Framework versions

• Transformers 4.46.1
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.20.1