wav2vec2-xls-r-2b-faroese-100h-60-epochs_v20250105_2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-2b on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1126
• Wer: 18.2932
• Cer: 3.9213

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 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
1.9282	0.4877	1000	0.4589	49.5572	14.5008
1.4639	0.9754	2000	0.2533	32.4360	8.4524
1.065	1.4628	3000	0.2168	30.5811	7.8859
0.7834	1.9505	4000	0.2123	30.1405	7.7542
0.6194	2.4379	5000	0.2405	31.5328	8.2552
0.6979	2.9256	6000	0.2070	29.9247	7.5964
0.6727	3.4131	7000	0.2236	31.0393	8.2149
0.6106	3.9008	8000	0.2121	29.8145	7.6989
0.5888	4.3882	9000	0.1929	28.8673	7.3328
0.5875	4.8759	10000	0.1903	28.4928	7.3076
0.5104	5.3633	11000	0.1922	28.6117	7.1624
0.4981	5.8510	12000	0.1713	27.4662	6.8713
0.4147	6.3385	13000	0.1799	27.0036	6.6835
0.4442	6.8261	14000	0.1768	27.6600	6.9904
0.3889	7.3136	15000	0.1729	26.5850	6.5620
0.3941	7.8013	16000	0.1670	26.7172	6.7435
0.3325	8.2887	17000	0.1763	26.5101	6.5991
0.3761	8.7764	18000	0.1591	25.9506	6.4555
0.2741	9.2638	19000	0.1674	25.4615	6.2219
0.2918	9.7515	20000	0.1638	25.8757	6.3600
0.2684	10.2390	21000	0.1621	25.0650	6.1217
0.2714	10.7267	22000	0.1443	24.8227	6.0239
0.2889	11.2141	23000	0.1512	24.6817	6.0176
0.2511	11.7018	24000	0.1506	24.1309	5.8125
0.2377	12.1892	25000	0.1513	23.9062	5.8054
0.2285	12.6769	26000	0.1560	24.1706	5.8701
0.2115	13.1644	27000	0.1557	23.9811	5.8330
0.228	13.6520	28000	0.1411	23.5978	5.6602
0.1855	14.1395	29000	0.1444	23.5097	5.5623
0.2058	14.6272	30000	0.1488	23.7388	5.6910
0.1894	15.1146	31000	0.1568	23.4084	5.6310
0.1882	15.6023	32000	0.1432	23.2321	5.5434
0.1794	16.0897	33000	0.1454	22.8576	5.4937
0.1535	16.5774	34000	0.1491	22.8753	5.3888
0.1688	17.0649	35000	0.1390	22.9634	5.4645
0.1705	17.5525	36000	0.1459	22.9281	5.3998
0.1401	18.0400	37000	0.1466	22.8709	5.4330
0.1476	18.5277	38000	0.1440	22.8841	5.3698
0.1426	19.0151	39000	0.1436	22.5933	5.2799
0.1375	19.5028	40000	0.1431	22.2584	5.2428
0.1491	19.9905	41000	0.1339	22.1307	5.1110
0.1564	20.4779	42000	0.1303	21.8619	5.0771
0.1427	20.9656	43000	0.1392	21.9809	5.1710
0.1145	21.4531	44000	0.1332	21.7518	5.0661
0.1453	21.9407	45000	0.1289	21.8839	5.0953
0.1184	22.4282	46000	0.1348	21.8707	5.0811
0.1349	22.9159	47000	0.1316	21.6240	4.9722
0.1196	23.4033	48000	0.1315	21.8575	5.0393
0.1199	23.8910	49000	0.1323	21.7430	5.0692
0.1177	24.3784	50000	0.1356	21.5579	4.9738
0.1101	24.8661	51000	0.1343	21.6108	5.0187
0.1173	25.3536	52000	0.1286	21.4125	4.9327
0.1256	25.8413	53000	0.1283	21.4257	4.8878
0.1157	26.3287	54000	0.1340	21.1570	4.8428
0.1044	26.8164	55000	0.1277	21.2143	4.8460
0.1019	27.3038	56000	0.1299	21.4346	4.9217
0.1116	27.7915	57000	0.1270	21.1217	4.8239
0.1055	28.2790	58000	0.1323	21.2231	4.8278
0.1047	28.7666	59000	0.1231	20.9191	4.7315
0.1129	29.2541	60000	0.1337	21.1394	4.8112
0.1174	29.7418	61000	0.1319	20.7560	4.7481
0.0992	30.2292	62000	0.1342	21.0777	4.8538
0.1128	30.7169	63000	0.1297	21.0292	4.7781
0.1103	31.2043	64000	0.1282	20.7428	4.6850
0.1076	31.6920	65000	0.1272	20.5710	4.6866
0.0998	32.1795	66000	0.1329	20.5974	4.6708
0.0863	32.6672	67000	0.1186	20.5005	4.5998
0.1026	33.1546	68000	0.1225	20.6635	4.6763
0.0782	33.6423	69000	0.1345	20.5622	4.6913
0.0837	34.1297	70000	0.1349	20.5137	4.6069
0.1144	34.6174	71000	0.1220	20.4873	4.6511
0.0969	35.1049	72000	0.1240	20.5886	4.6353
0.0848	35.5925	73000	0.1307	20.2934	4.5911
0.0748	36.0800	74000	0.1275	20.2229	4.5201
0.0929	36.5677	75000	0.1184	20.1745	4.4909
0.0877	37.0551	76000	0.1287	20.0687	4.5106
0.0932	37.5428	77000	0.1189	20.0599	4.5233
0.0923	38.0302	78000	0.1303	20.1877	4.5138
0.068	38.5179	79000	0.1274	19.8881	4.4412
0.1069	39.0054	80000	0.1219	20.0952	4.4956
0.0735	39.4931	81000	0.1225	19.8572	4.4420
0.0745	39.9807	82000	0.1246	19.7339	4.4065
0.0664	40.4682	83000	0.1262	19.8132	4.4104
0.0783	40.9559	84000	0.1285	19.8000	4.4664
0.0884	41.4433	85000	0.1123	19.5312	4.3181
0.0704	41.9310	86000	0.1226	19.5356	4.3434
0.0675	42.4184	87000	0.1330	19.7030	4.3536
0.093	42.9061	88000	0.1207	19.5841	4.3662
0.0809	43.3936	89000	0.1217	19.4872	4.3142
0.0668	43.8812	90000	0.1185	19.5180	4.2731
0.0968	44.3687	91000	0.1156	19.4079	4.2818
0.0568	44.8564	92000	0.1191	19.5180	4.2992
0.0731	45.3438	93000	0.1165	19.4695	4.2653
0.0562	45.8315	94000	0.1220	19.4431	4.2692
0.064	46.3189	95000	0.1213	19.3946	4.2589
0.0638	46.8066	96000	0.1202	19.0906	4.1682
0.0509	47.2941	97000	0.1232	19.2977	4.2290
0.0585	47.7818	98000	0.1147	19.2052	4.1935
0.0592	48.2692	99000	0.1199	19.1038	4.1556
0.0488	48.7569	100000	0.1159	19.0862	4.1580
0.0513	49.2443	101000	0.1168	19.0510	4.1185
0.076	49.7320	102000	0.1053	18.8351	4.0743
0.0466	50.2195	103000	0.1129	18.8836	4.1035
0.0472	50.7071	104000	0.1210	18.8836	4.1114
0.0433	51.1946	105000	0.1154	18.7822	4.0751
0.056	51.6823	106000	0.1161	18.7734	4.0854
0.0419	52.1697	107000	0.1125	18.6456	4.0365
0.0432	52.6574	108000	0.1113	18.6677	4.0167
0.0438	53.1448	109000	0.1139	18.5840	4.0136
0.0387	53.6325	110000	0.1119	18.6104	4.0207
0.0356	54.1200	111000	0.1114	18.5575	3.9757
0.0365	54.6077	112000	0.1187	18.5311	3.9844
0.0416	55.0951	113000	0.1136	18.5135	3.9986
0.0456	55.5828	114000	0.1179	18.4518	3.9749
0.0249	56.0702	115000	0.1123	18.3769	3.9370
0.0324	56.5579	116000	0.1148	18.3945	3.9410
0.0383	57.0454	117000	0.1127	18.3108	3.9307
0.0287	57.5330	118000	0.1100	18.4077	3.9370
0.026	58.0205	119000	0.1122	18.3460	3.9228
0.0284	58.5082	120000	0.1137	18.3372	3.9299
0.0287	58.9959	121000	0.1130	18.2755	3.9149
0.0283	59.4833	122000	0.1122	18.2844	3.9197
0.0286	59.9710	123000	0.1126	18.2932	3.9213

Framework versions

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