wav2vec2-xls-r-300m-CV_Fleurs_AMMI_ALFFA-sw-10hrs-v1

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: 1.2971
• Wer: 0.3556
• Cer: 0.1044

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: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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
7.0482	0.9986	360	2.9509	1.0	1.0
5.5647	2.0	721	2.1843	1.0	0.6661
3.4187	2.9986	1081	1.3719	0.9273	0.3550
2.5943	4.0	1442	1.1633	0.8977	0.2827
2.1874	4.9986	1802	0.9978	0.7963	0.2396
1.8944	6.0	2163	0.8786	0.7200	0.2105
1.6523	6.9986	2523	0.8251	0.6626	0.1915
1.4683	8.0	2884	0.7958	0.6460	0.1890
1.3358	8.9986	3244	0.7706	0.6021	0.1754
1.1904	10.0	3605	0.7385	0.5992	0.1751
1.0935	10.9986	3965	0.7722	0.5620	0.1655
1.0084	12.0	4326	0.7509	0.5715	0.1697
0.9096	12.9986	4686	0.7622	0.5253	0.1503
0.8377	14.0	5047	0.7561	0.5150	0.1496
0.7773	14.9986	5407	0.7909	0.5108	0.1511
0.7337	16.0	5768	0.7987	0.5079	0.1481
0.6904	16.9986	6128	0.8391	0.4989	0.1426
0.6381	18.0	6489	0.8506	0.5009	0.1422
0.6163	18.9986	6849	0.8754	0.4815	0.1392
0.5792	20.0	7210	0.9556	0.4788	0.1373
0.5665	20.9986	7570	0.8177	0.4739	0.1371
0.5387	22.0	7931	0.8381	0.4704	0.1375
0.5125	22.9986	8291	0.8817	0.4589	0.1323
0.4893	24.0	8652	0.9227	0.4590	0.1344
0.4702	24.9986	9012	1.0008	0.4470	0.1294
0.4583	26.0	9373	0.9699	0.4488	0.1298
0.448	26.9986	9733	1.0021	0.4525	0.1320
0.4358	28.0	10094	0.9996	0.4459	0.1313
0.42	28.9986	10454	0.9673	0.4434	0.1273
0.3913	30.0	10815	1.0123	0.4410	0.1272
0.4023	30.9986	11175	1.0258	0.4308	0.1258
0.3793	32.0	11536	1.0015	0.4334	0.1269
0.3716	32.9986	11896	1.0227	0.4370	0.1273
0.3668	34.0	12257	1.0355	0.4283	0.1251
0.3442	34.9986	12617	1.0020	0.4264	0.1235
0.3392	36.0	12978	1.0422	0.4214	0.1238
0.3229	36.9986	13338	1.0545	0.4219	0.1236
0.311	38.0	13699	1.0842	0.4227	0.1242
0.3046	38.9986	14059	1.0755	0.4168	0.1208
0.2966	40.0	14420	1.0766	0.4206	0.1239
0.289	40.9986	14780	1.0696	0.4119	0.1209
0.2801	42.0	15141	1.0653	0.4127	0.1215
0.2845	42.9986	15501	1.1241	0.4138	0.1215
0.2795	44.0	15862	1.1176	0.4077	0.1195
0.2742	44.9986	16222	1.0823	0.4092	0.1198
0.2772	46.0	16583	1.0684	0.4111	0.1214
0.2607	46.9986	16943	1.1543	0.4055	0.1191
0.2565	48.0	17304	1.1207	0.4042	0.1186
0.2397	48.9986	17664	1.1324	0.4023	0.1184
0.2439	50.0	18025	1.0794	0.4058	0.1192
0.2393	50.9986	18385	1.1105	0.4002	0.1177
0.2283	52.0	18746	1.1724	0.3938	0.1162
0.226	52.9986	19106	1.1660	0.3975	0.1175
0.2272	54.0	19467	1.1530	0.3959	0.1162
0.2169	54.9986	19827	1.1406	0.3951	0.1150
0.2186	56.0	20188	1.1512	0.3952	0.1164
0.2221	56.9986	20548	1.1636	0.3945	0.1159
0.2113	58.0	20909	1.1598	0.3935	0.1151
0.2033	58.9986	21269	1.1667	0.3929	0.1154
0.201	60.0	21630	1.1330	0.3902	0.1142
0.2049	60.9986	21990	1.1746	0.3905	0.1138
0.2043	62.0	22351	1.1990	0.3869	0.1133
0.1945	62.9986	22711	1.1931	0.3900	0.1138
0.1876	64.0	23072	1.1713	0.3833	0.1131
0.1842	64.9986	23432	1.1645	0.3832	0.1127
0.1842	66.0	23793	1.1445	0.3844	0.1129
0.1853	66.9986	24153	1.1917	0.3853	0.1125
0.1745	68.0	24514	1.1785	0.3796	0.1121
0.1703	68.9986	24874	1.1752	0.3815	0.1114
0.1711	70.0	25235	1.1820	0.3784	0.1107
0.1687	70.9986	25595	1.1864	0.3788	0.1105
0.1632	72.0	25956	1.2335	0.3758	0.1097
0.1613	72.9986	26316	1.1948	0.3753	0.1099
0.1551	74.0	26677	1.2681	0.3767	0.1098
0.1606	74.9986	27037	1.2648	0.3738	0.1098
0.1445	76.0	27398	1.2243	0.3739	0.1089
0.1562	76.9986	27758	1.2221	0.3718	0.1089
0.1457	78.0	28119	1.2586	0.3709	0.1081
0.1431	78.9986	28479	1.2694	0.3731	0.1093
0.138	80.0	28840	1.2734	0.3707	0.1092
0.1382	80.9986	29200	1.2558	0.3668	0.1078
0.1389	82.0	29561	1.2251	0.3683	0.1088
0.1365	82.9986	29921	1.2232	0.3663	0.1081
0.13	84.0	30282	1.2646	0.3663	0.1079
0.1302	84.9986	30642	1.2837	0.3649	0.1073
0.1262	86.0	31003	1.2755	0.3636	0.1063
0.1276	86.9986	31363	1.2917	0.3637	0.1068
0.1217	88.0	31724	1.2952	0.3623	0.1064
0.1228	88.9986	32084	1.2751	0.3602	0.1057
0.1196	90.0	32445	1.2764	0.3606	0.1056
0.1214	90.9986	32805	1.2727	0.3603	0.1057
0.1212	92.0	33166	1.2687	0.3582	0.1051
0.1157	92.9986	33526	1.2731	0.3576	0.1050
0.1134	94.0	33887	1.2842	0.3580	0.1052
0.1119	94.9986	34247	1.3028	0.3572	0.1051
0.1132	96.0	34608	1.2819	0.3562	0.1047
0.1117	96.9986	34968	1.2993	0.3561	0.1047
0.1078	98.0	35329	1.3051	0.3547	0.1045
0.1088	98.9986	35689	1.2980	0.3554	0.1043
0.1066	99.8613	36000	1.2971	0.3556	0.1044

Framework versions

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