wav2vec2-xls-r-1b-scandinavian-faroese-100h-60-epochs-20250112_v4

This model is a fine-tuned version of davidilag/wav2vec2-xls-r-1b-scandinavian-251h-30-epochs-20250111_v9 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1070
• Wer: 17.8261
• Cer: 3.7769

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 6000
• num_epochs: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.0345	0.4877	1000	0.3260	39.9656	10.9180
0.7814	0.9754	2000	0.2024	29.6867	7.6587
0.6454	1.4628	3000	0.1783	27.9376	7.1064
0.6608	1.9505	4000	0.1803	28.9377	7.2713
0.5288	2.4379	5000	0.2009	28.5060	7.2681
0.6023	2.9256	6000	0.1948	29.0787	7.4425
0.5827	3.4131	7000	0.1799	28.5544	7.2760
0.5251	3.9008	8000	0.1759	27.4794	6.9036
0.489	4.3882	9000	0.1710	26.3647	6.7032
0.5095	4.8759	10000	0.1579	26.5013	6.6078
0.446	5.3633	11000	0.1476	25.0518	6.1541
0.4329	5.8510	12000	0.1504	25.1928	6.2038
0.3607	6.3385	13000	0.1389	24.7125	5.9892
0.3411	6.8261	14000	0.1480	24.7478	6.1312
0.3508	7.3136	15000	0.1381	24.3116	5.8511
0.3302	7.8013	16000	0.1430	24.5936	5.9576
0.2681	8.2887	17000	0.1434	23.8093	5.7012
0.2914	8.7764	18000	0.1312	23.8225	5.7099
0.2386	9.2638	19000	0.1334	23.4436	5.5766
0.2496	9.7515	20000	0.1363	23.6507	5.6207
0.2329	10.2390	21000	0.1306	23.0515	5.3754
0.2416	10.7267	22000	0.1300	22.9986	5.4779
0.2566	11.2141	23000	0.1235	22.6990	5.3335
0.2267	11.7018	24000	0.1301	22.8356	5.3659
0.2147	12.1892	25000	0.1276	22.4743	5.2657
0.2009	12.6769	26000	0.1246	22.2717	5.2136
0.1959	13.1644	27000	0.1214	22.0293	5.1434
0.2053	13.6520	28000	0.1211	22.0866	5.0890
0.1648	14.1395	29000	0.1309	21.8707	5.0361
0.1711	14.6272	30000	0.1265	22.0249	5.0850
0.1612	15.1146	31000	0.1214	21.7165	4.9848
0.162	15.6023	32000	0.1210	21.7209	5.0030
0.1483	16.0897	33000	0.1265	21.5888	4.9935
0.1539	16.5774	34000	0.1203	21.6020	4.9801
0.1382	17.0649	35000	0.1139	21.1966	4.8207
0.159	17.5525	36000	0.1154	21.2099	4.8128
0.1176	18.0400	37000	0.1213	21.2715	4.8010
0.1218	18.5277	38000	0.1175	20.9411	4.7860
0.1239	19.0151	39000	0.1195	20.9719	4.7742
0.1277	19.5028	40000	0.1133	21.0865	4.7410
0.1327	19.9905	41000	0.1117	20.7120	4.6842
0.1403	20.4779	42000	0.1247	21.0645	4.8128
0.1251	20.9656	43000	0.1083	20.6591	4.6416
0.1084	21.4531	44000	0.1174	20.8001	4.6795
0.1258	21.9407	45000	0.1185	20.6371	4.6495
0.0906	22.4282	46000	0.1232	20.6723	4.6740
0.1074	22.9159	47000	0.1187	20.3771	4.6416
0.1125	23.4033	48000	0.1138	20.4212	4.6140
0.1055	23.8910	49000	0.1220	20.5754	4.7008
0.1116	24.3784	50000	0.1181	20.3463	4.6100
0.0997	24.8661	51000	0.1261	20.3331	4.6361
0.1025	25.3536	52000	0.1166	20.2714	4.6085
0.1032	25.8413	53000	0.1118	20.1833	4.4862
0.0997	26.3287	54000	0.1165	20.2978	4.5603
0.0923	26.8164	55000	0.1121	20.0555	4.4901
0.087	27.3038	56000	0.1236	20.2229	4.5785
0.0961	27.7915	57000	0.1130	19.8881	4.4380
0.095	28.2790	58000	0.1168	19.9806	4.4570
0.0995	28.7666	59000	0.1187	20.1436	4.5501
0.1021	29.2541	60000	0.1195	20.1613	4.5209
0.1168	29.7418	61000	0.1211	19.8132	4.4412
0.1029	30.2292	62000	0.1161	19.9277	4.3994
0.1023	30.7169	63000	0.1148	19.7868	4.4089
0.1044	31.2043	64000	0.1089	19.5753	4.3599
0.0886	31.6920	65000	0.1109	19.4211	4.3165
0.0885	32.1795	66000	0.1192	19.5621	4.3576
0.0723	32.6672	67000	0.1170	19.4960	4.3505
0.078	33.1546	68000	0.1129	19.5444	4.3292
0.066	33.6423	69000	0.1197	19.4255	4.2960
0.0684	34.1297	70000	0.1205	19.7603	4.3402
0.0856	34.6174	71000	0.1131	19.5180	4.3173
0.0857	35.1049	72000	0.1141	19.6590	4.3473
0.0813	35.5925	73000	0.1168	19.4828	4.2881
0.0752	36.0800	74000	0.1181	19.3594	4.2684
0.0722	36.5677	75000	0.1112	19.2096	4.2313
0.0699	37.0551	76000	0.1178	19.1787	4.2203
0.0717	37.5428	77000	0.1112	19.0334	4.1903
0.1025	38.0302	78000	0.1081	19.1171	4.1911
0.0668	38.5179	79000	0.1176	19.0334	4.1745
0.0858	39.0054	80000	0.1089	19.0201	4.1580
0.078	39.4931	81000	0.1079	19.0466	4.1777
0.0718	39.9807	82000	0.1079	19.0950	4.1390
0.0766	40.4682	83000	0.1109	19.0025	4.1335
0.0611	40.9559	84000	0.1136	18.8747	4.1146
0.0846	41.4433	85000	0.1093	18.8703	4.1153
0.0719	41.9310	86000	0.1107	18.8924	4.1106
0.0551	42.4184	87000	0.1077	18.9540	4.0625
0.0849	42.9061	88000	0.1026	18.8747	4.0846
0.0715	43.3936	89000	0.1106	18.8042	4.0830
0.0682	43.8812	90000	0.1157	18.8659	4.0956
0.0754	44.3687	91000	0.1137	18.6677	4.0467
0.0627	44.8564	92000	0.1154	18.6809	4.0341
0.0821	45.3438	93000	0.1076	18.5972	4.0459
0.0514	45.8315	94000	0.1088	18.5619	4.0175
0.0505	46.3189	95000	0.1110	18.5972	4.0049
0.0611	46.8066	96000	0.1123	18.5179	4.0270
0.0568	47.2941	97000	0.1099	18.4209	3.9836
0.0464	47.7818	98000	0.1047	18.4606	3.9694
0.0543	48.2692	99000	0.1075	18.3460	3.9449
0.0464	48.7569	100000	0.1101	18.3020	3.9339
0.0408	49.2443	101000	0.1070	18.2006	3.9031
0.0632	49.7320	102000	0.1053	18.3196	3.9078
0.0648	50.2195	103000	0.1050	18.3593	3.9370
0.0479	50.7071	104000	0.1080	18.4253	3.9702
0.0534	51.1946	105000	0.1054	18.2183	3.9031
0.0361	51.6823	106000	0.1076	18.0685	3.8850
0.0402	52.1697	107000	0.1010	18.1478	3.8629
0.0367	52.6574	108000	0.1076	18.1125	3.8637
0.0465	53.1448	109000	0.1032	18.0244	3.8266
0.0364	53.6325	110000	0.1055	17.9936	3.8353
0.044	54.1200	111000	0.1041	18.0905	3.8637
0.0451	54.6077	112000	0.1060	18.0244	3.8384
0.0438	55.0951	113000	0.1078	17.9099	3.8155
0.045	55.5828	114000	0.1062	17.8658	3.8124
0.0259	56.0702	115000	0.1068	17.9143	3.7927
0.0393	56.5579	116000	0.1058	17.9055	3.7879
0.0457	57.0454	117000	0.1060	17.8746	3.7927
0.029	57.5330	118000	0.1055	17.8306	3.7761
0.0265	58.0205	119000	0.1067	17.8746	3.7856
0.0282	58.5082	120000	0.1073	17.8217	3.7737
0.0264	58.9959	121000	0.1076	17.8438	3.7792
0.0295	59.4833	122000	0.1072	17.8173	3.7721
0.0276	59.9710	123000	0.1070	17.8261	3.7769

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0