wav2vec2-xls-r-300m-cpt-200h-FO-IS-NO-DK-SE-cp-best-faroese-100h-30-epochs_run9_2025-09-11

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1019
• Wer: 18.8219
• Cer: 4.0508

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_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.2903	0.4877	1000	3.2182	100.0	99.7980
0.7571	0.9754	2000	0.4684	43.4859	11.9670
0.4077	1.4628	3000	0.2338	31.5108	7.9959
0.3474	1.9505	4000	0.2022	29.4709	7.3773
0.2736	2.4379	5000	0.1783	27.7702	6.8392
0.2543	2.9256	6000	0.1628	27.1534	6.6191
0.2051	3.4131	7000	0.1481	25.6862	6.1101
0.2105	3.9008	8000	0.1392	24.9901	5.9224
0.1741	4.3882	9000	0.1362	23.9855	5.6588
0.1836	4.8759	10000	0.1366	23.9767	5.6857
0.1477	5.3633	11000	0.1491	23.5406	5.5105
0.1544	5.8510	12000	0.1275	23.6243	5.5050
0.1346	6.3385	13000	0.1232	23.1793	5.3732
0.1337	6.8261	14000	0.1282	22.6858	5.2351
0.1161	7.3136	15000	0.1210	22.6065	5.2399
0.1248	7.8013	16000	0.1180	22.6726	5.1791
0.115	8.2887	17000	0.1174	22.0954	5.0663
0.1124	8.7764	18000	0.1174	22.0293	5.0213
0.0999	9.2638	19000	0.1145	21.6681	4.9700
0.1065	9.7515	20000	0.1179	21.4962	4.8935
0.0861	10.2390	21000	0.1190	21.3420	4.8580
0.0898	10.7267	22000	0.1162	21.3464	4.8446
0.0773	11.2141	23000	0.1182	21.1129	4.8043
0.0781	11.7018	24000	0.1195	21.1702	4.8351
0.0784	12.1892	25000	0.1067	20.7032	4.6623
0.0688	12.6769	26000	0.1146	20.8662	4.7033
0.0763	13.1644	27000	0.1081	20.7120	4.6386
0.0664	13.6520	28000	0.1124	20.6151	4.6339
0.065	14.1395	29000	0.1103	20.7428	4.6584
0.0753	14.6272	30000	0.1041	20.3595	4.5084
0.0624	15.1146	31000	0.1086	20.4564	4.5510
0.0566	15.6023	32000	0.1096	20.2229	4.4579
0.0638	16.0897	33000	0.1103	20.2934	4.4935
0.0611	16.5774	34000	0.1067	20.1084	4.4366
0.0506	17.0649	35000	0.1083	20.0379	4.4556
0.05	17.5525	36000	0.1008	19.7559	4.3372
0.0525	18.0400	37000	0.1034	19.7779	4.2978
0.0429	18.5277	38000	0.1121	19.6149	4.2970
0.0531	19.0151	39000	0.1042	19.7163	4.3278
0.0426	19.5028	40000	0.1085	19.6634	4.3009
0.0405	19.9905	41000	0.1066	19.5400	4.2670
0.0427	20.4779	42000	0.1074	19.4255	4.2149
0.0355	20.9656	43000	0.1019	19.4211	4.1818
0.0363	21.4531	44000	0.1053	19.3594	4.1873
0.054	21.9407	45000	0.1034	19.2625	4.1778
0.0416	22.4282	46000	0.1012	19.1215	4.1431
0.0432	22.9159	47000	0.1047	19.1611	4.1510
0.0432	23.4033	48000	0.1025	19.0378	4.1329
0.0364	23.8910	49000	0.1041	19.1567	4.1329
0.039	24.3784	50000	0.1043	19.1127	4.1210
0.0372	24.8661	51000	0.1042	18.8747	4.0871
0.032	25.3536	52000	0.1035	18.8703	4.0658
0.0323	25.8413	53000	0.1018	18.8395	4.0461
0.0365	26.3287	54000	0.1011	18.9276	4.0682
0.0297	26.8164	55000	0.1012	18.8131	4.0516
0.0363	27.3038	56000	0.1018	18.8395	4.0642
0.0382	27.7915	57000	0.1009	18.8087	4.0477
0.0379	28.2790	58000	0.1017	18.8307	4.0500
0.0301	28.7666	59000	0.1017	18.8087	4.0492
0.0397	29.2541	60000	0.1019	18.8175	4.0516
0.0365	29.7418	61000	0.1019	18.8219	4.0508

Framework versions

• Transformers 4.56.1
• Pytorch 2.8.0+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0