wav2vec2-xls-r-300m-Fleurs_AMMI_AFRIVOICE_LRSC-ln-5hrs-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: 0.4635
• Wer: 0.3074
• Cer: 0.0975

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
10.1117	1.0	394	4.3490	1.0	1.0
3.7471	2.0	788	3.1184	1.0	1.0
3.0204	3.0	1182	2.9072	1.0	1.0
2.9064	4.0	1576	2.8651	1.0	1.0
2.8666	5.0	1970	2.8332	1.0	1.0
2.8284	6.0	2364	2.8052	1.0	1.0
2.7745	7.0	2758	2.6265	1.0	0.9998
2.4996	8.0	3152	2.0214	1.0	0.5429
1.9287	9.0	3546	1.4032	0.9991	0.3316
1.5136	10.0	3940	1.1202	0.9125	0.2549
1.2999	11.0	4334	0.9748	0.7513	0.2049
1.1427	12.0	4728	0.8717	0.6135	0.1733
1.0469	13.0	5122	0.8070	0.5762	0.1658
0.9806	14.0	5516	0.7431	0.5455	0.1568
0.9187	15.0	5910	0.7146	0.5244	0.1512
0.8663	16.0	6304	0.6732	0.4882	0.1419
0.8084	17.0	6698	0.6555	0.4738	0.1390
0.7852	18.0	7092	0.6295	0.4550	0.1350
0.7505	19.0	7486	0.6140	0.4433	0.1311
0.7193	20.0	7880	0.6098	0.4250	0.1268
0.6927	21.0	8274	0.5722	0.4215	0.1260
0.6564	22.0	8668	0.5742	0.4074	0.1217
0.6699	23.0	9062	0.5864	0.4032	0.1220
0.6269	24.0	9456	0.5620	0.3918	0.1189
0.6142	25.0	9850	0.5484	0.3901	0.1177
0.6001	26.0	10244	0.5475	0.3829	0.1164
0.5953	27.0	10638	0.5361	0.3700	0.1132
0.584	28.0	11032	0.5283	0.3718	0.1136
0.5555	29.0	11426	0.5323	0.3693	0.1129
0.5453	30.0	11820	0.5048	0.3665	0.1126
0.5363	31.0	12214	0.5200	0.3542	0.1093
0.5196	32.0	12608	0.5118	0.3512	0.1086
0.5098	33.0	13002	0.5059	0.3514	0.1087
0.5042	34.0	13396	0.5109	0.3531	0.1094
0.4955	35.0	13790	0.4946	0.3456	0.1080
0.4958	36.0	14184	0.4943	0.3399	0.1069
0.4743	37.0	14578	0.4953	0.3395	0.1065
0.4739	38.0	14972	0.4996	0.3432	0.1077
0.4616	39.0	15366	0.4896	0.3393	0.1051
0.4509	40.0	15760	0.4793	0.3336	0.1042
0.4523	41.0	16154	0.4955	0.3316	0.1037
0.4301	42.0	16548	0.4827	0.3384	0.1047
0.4312	43.0	16942	0.4763	0.3247	0.1016
0.4157	44.0	17336	0.4786	0.3244	0.1016
0.4177	45.0	17730	0.4948	0.3273	0.1025
0.4207	46.0	18124	0.4869	0.3202	0.1016
0.4028	47.0	18518	0.4852	0.3259	0.1030
0.41	48.0	18912	0.4745	0.3231	0.1015
0.4042	49.0	19306	0.4866	0.3191	0.1003
0.3953	50.0	19700	0.4785	0.3111	0.0975
0.3889	51.0	20094	0.4798	0.3196	0.0992
0.3853	52.0	20488	0.4858	0.3182	0.0993
0.3819	53.0	20882	0.4848	0.3179	0.1005
0.3761	54.0	21276	0.4759	0.3134	0.0992
0.3652	55.0	21670	0.4643	0.3160	0.0996
0.3734	56.0	22064	0.4868	0.3062	0.0975
0.3574	57.0	22458	0.4636	0.3105	0.0989
0.3538	58.0	22852	0.4596	0.3139	0.1002
0.3731	59.0	23246	0.4738	0.3150	0.0996
0.3497	60.0	23640	0.4609	0.3152	0.0994
0.3393	61.0	24034	0.4627	0.3063	0.0975
0.3557	62.0	24428	0.4860	0.3032	0.0959
0.3461	63.0	24822	0.4818	0.3059	0.0975
0.3395	64.0	25216	0.4809	0.3049	0.0969
0.3314	65.0	25610	0.4736	0.3039	0.0958
0.3372	66.0	26004	0.4698	0.3066	0.0974
0.3356	67.0	26398	0.4809	0.2984	0.0949
0.3269	68.0	26792	0.4627	0.3074	0.0969
0.3206	69.0	27186	0.4681	0.3005	0.0966
0.3282	70.0	27580	0.4734	0.3031	0.0965
0.3256	71.0	27974	0.4781	0.3006	0.0952
0.3187	72.0	28368	0.4684	0.3061	0.0973
0.3229	73.0	28762	0.4793	0.2988	0.0953
0.3068	74.0	29156	0.4662	0.3010	0.0961
0.318	75.0	29550	0.4826	0.2996	0.0957
0.3053	76.0	29944	0.4695	0.2986	0.0946
0.3117	77.0	30338	0.4635	0.3074	0.0975

Framework versions

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