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: cosine
lr_scheduler_warmup_steps: 5000
num_epochs: 30
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.0663	0.4877	1000	0.4799	54.6284	15.5160
0.5134	0.9754	2000	0.2491	33.6080	8.6429
0.4176	1.4628	3000	0.2075	31.1187	7.9399
0.3923	1.9505	4000	0.2237	31.8632	8.1789
0.3497	2.4379	5000	0.2379	33.6388	8.6421
0.3539	2.9256	6000	0.2181	31.1715	8.1561
0.3228	3.4131	7000	0.2242	31.8016	8.2728
0.295	3.9008	8000	0.2056	30.0656	7.6677
0.2754	4.3882	9000	0.1886	28.7527	7.3079
0.2802	4.8759	10000	0.1936	28.7527	7.3607
0.2589	5.3633	11000	0.1863	27.8627	7.0507
0.2335	5.8510	12000	0.1745	27.9464	7.1193
0.1956	6.3385	13000	0.1745	27.5235	6.7982
0.2026	6.8261	14000	0.1618	26.7657	6.7532
0.1986	7.3136	15000	0.1658	26.6599	6.5962
0.1901	7.8013	16000	0.1740	26.7040	6.5820
0.1621	8.2887	17000	0.1668	25.8933	6.3958
0.1651	8.7764	18000	0.1534	25.6245	6.2979
0.1315	9.2638	19000	0.1583	25.3514	6.2411
0.1545	9.7515	20000	0.1635	24.7830	6.1228
0.1361	10.2390	21000	0.1524	25.3470	6.2443
0.126	10.7267	22000	0.1495	24.6200	5.9516
0.1291	11.2141	23000	0.1419	24.0913	5.8892
0.1212	11.7018	24000	0.1390	23.7961	5.6667
0.1097	12.1892	25000	0.1435	24.1089	5.7030
0.1132	12.6769	26000	0.1391	23.8446	5.6494
0.0948	13.1644	27000	0.1537	23.5273	5.6210
0.0934	13.6520	28000	0.1445	23.1264	5.4876
0.0854	14.1395	29000	0.1333	22.6153	5.2343
0.0878	14.6272	30000	0.1315	22.6638	5.3787
0.0805	15.1146	31000	0.1304	22.4964	5.2075
0.074	15.6023	32000	0.1295	22.4303	5.1602
0.0734	16.0897	33000	0.1279	22.2937	5.1404
0.0692	16.5774	34000	0.1234	22.0602	5.0773
0.0623	17.0649	35000	0.1309	21.7914	5.0284
0.0693	17.5525	36000	0.1233	21.6328	4.9448
0.057	18.0400	37000	0.1225	21.5050	4.8935
0.0479	18.5277	38000	0.1238	21.3288	4.8722
0.0478	19.0151	39000	0.1285	21.2715	4.8406
0.0483	19.5028	40000	0.1228	20.9499	4.7191
0.0435	19.9905	41000	0.1217	20.7516	4.6489
0.052	20.4779	42000	0.1170	20.5269	4.5795
0.0436	20.9656	43000	0.1179	20.5137	4.5913
0.0388	21.4531	44000	0.1196	20.3639	4.5211
0.0464	21.9407	45000	0.1123	20.2009	4.4800
0.0321	22.4282	46000	0.1151	20.0819	4.4619
0.0354	22.9159	47000	0.1143	19.8881	4.3893
0.033	23.4033	48000	0.1104	20.0555	4.4145
0.0315	23.8910	49000	0.1113	19.8352	4.3096
0.0301	24.3784	50000	0.1149	19.7207	4.3104
0.0272	24.8661	51000	0.1141	19.4563	4.2236
0.0323	25.3536	52000	0.1155	19.5092	4.2441
0.032	25.8413	53000	0.1115	19.4211	4.2062
0.0282	26.3287	54000	0.1113	19.3109	4.1613
0.0257	26.8164	55000	0.1118	19.3197	4.1542
0.023	27.3038	56000	0.1121	19.2581	4.1463
0.0305	27.7915	57000	0.1122	19.1743	4.1384
0.0287	28.2790	58000	0.1123	19.1964	4.1337
0.0252	28.7666	59000	0.1121	19.2316	4.1344
0.0325	29.2541	60000	0.1115	19.2228	4.1376
0.0374	29.7418	61000	0.1114	19.2228	4.1352

Framework versions

Transformers 4.48.1
Pytorch 2.5.1+cu124
Datasets 3.2.0
Tokenizers 0.21.0

davidilag/wav2vec2-xls-r-2b-faroese-100h-30-epochs_20250126

wav2vec2-xls-r-2b-faroese-100h-30-epochs_20250126

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

davidilag/
wav2vec2-xls-r-2b-faroese-100h-30-epochs_20250126