wav2vec2-xls-r-300m-faroese-100h-60-epochs_20250122

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1124
• Wer: 18.4782
• Cer: 3.9149

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
3.4358	0.4877	1000	3.3478	100.0	98.6201
1.7182	0.9754	2000	1.2803	85.5267	29.7266
0.6446	1.4628	3000	0.3541	42.8471	11.2580
0.5175	1.9505	4000	0.2806	37.2869	9.5388
0.3949	2.4379	5000	0.2186	32.0659	8.0058
0.3617	2.9256	6000	0.1976	30.4710	7.5987
0.2844	3.4131	7000	0.1740	28.1799	6.9052
0.2947	3.9008	8000	0.1591	27.3781	6.5904
0.242	4.3882	9000	0.1474	26.4044	6.2251
0.2464	4.8759	10000	0.1466	26.0651	6.2669
0.1967	5.3633	11000	0.1528	25.1267	5.9774
0.2109	5.8510	12000	0.1286	24.7962	5.8574
0.1707	6.3385	13000	0.1310	24.8183	5.8022
0.1799	6.8261	14000	0.1284	23.7917	5.5505
0.1513	7.3136	15000	0.1294	23.7564	5.5813
0.1585	7.8013	16000	0.1195	23.3335	5.4006
0.1505	8.2887	17000	0.1214	23.1572	5.3785
0.1479	8.7764	18000	0.1291	22.9017	5.3130
0.1309	9.2638	19000	0.1229	22.7916	5.3328
0.1417	9.7515	20000	0.1210	22.5140	5.2184
0.1292	10.2390	21000	0.1166	22.3730	5.1047
0.118	10.7267	22000	0.1153	22.4479	5.1371
0.1042	11.2141	23000	0.1248	22.2144	5.0858
0.098	11.7018	24000	0.1189	22.0161	4.9990
0.1063	12.1892	25000	0.1182	21.8399	5.0408
0.0986	12.6769	26000	0.1225	21.8972	4.9935
0.1026	13.1644	27000	0.1130	21.5403	4.8538
0.0968	13.6520	28000	0.1151	21.5888	4.9059
0.0889	14.1395	29000	0.1155	21.5888	4.9564
0.1088	14.6272	30000	0.1071	21.2319	4.7970
0.0882	15.1146	31000	0.1157	21.5491	4.8341
0.0858	15.6023	32000	0.1141	21.2407	4.8073
0.0896	16.0897	33000	0.1122	20.9367	4.7102
0.0849	16.5774	34000	0.1147	20.9675	4.7710
0.0719	17.0649	35000	0.1149	21.2848	4.7947
0.0694	17.5525	36000	0.1263	20.9411	4.7489
0.0785	18.0400	37000	0.1114	20.8750	4.6597
0.0722	18.5277	38000	0.1146	20.9763	4.7016
0.0748	19.0151	39000	0.1101	20.7208	4.6290
0.0676	19.5028	40000	0.1120	20.7737	4.6511
0.067	19.9905	41000	0.1120	20.5049	4.5722
0.0681	20.4779	42000	0.1232	21.0645	4.6637
0.0638	20.9656	43000	0.1096	20.5798	4.5446
0.0574	21.4531	44000	0.1137	20.7208	4.6069
0.0874	21.9407	45000	0.1155	20.5049	4.5738
0.0716	22.4282	46000	0.1118	20.3331	4.5335
0.078	22.9159	47000	0.1140	20.5181	4.5769
0.0692	23.4033	48000	0.1174	20.3551	4.5272
0.0582	23.8910	49000	0.1229	20.2758	4.4941
0.0677	24.3784	50000	0.1108	20.0864	4.4467
0.0638	24.8661	51000	0.1119	20.1216	4.4451
0.0591	25.3536	52000	0.1120	20.3419	4.4657
0.0538	25.8413	53000	0.1147	19.9586	4.4120
0.0638	26.3287	54000	0.1104	20.0159	4.3883
0.0503	26.8164	55000	0.1160	20.0247	4.4160
0.0614	27.3038	56000	0.1124	20.0070	4.3868
0.0632	27.7915	57000	0.1165	20.0335	4.4160
0.0686	28.2790	58000	0.1109	20.0819	4.3741
0.0518	28.7666	59000	0.1112	19.8352	4.3733
0.0595	29.2541	60000	0.1106	19.7251	4.3655
0.0583	29.7418	61000	0.1120	19.7163	4.3710
0.0596	30.2292	62000	0.1111	19.7779	4.3268
0.0703	30.7169	63000	0.1117	19.9277	4.3812
0.058	31.2043	64000	0.1161	19.8572	4.3655
0.0577	31.6920	65000	0.1041	19.5841	4.2440
0.0638	32.1795	66000	0.1191	19.7735	4.3268
0.0596	32.6672	67000	0.1110	19.7779	4.3118
0.0569	33.1546	68000	0.1140	19.6766	4.2771
0.0528	33.6423	69000	0.1128	19.7030	4.3039
0.0488	34.1297	70000	0.1152	19.5532	4.2668
0.0635	34.6174	71000	0.1086	19.5048	4.2463
0.0574	35.1049	72000	0.1137	19.5621	4.2503
0.0473	35.5925	73000	0.1111	19.5885	4.2558
0.0442	36.0800	74000	0.1182	19.4872	4.2392
0.0504	36.5677	75000	0.1111	19.4431	4.2219
0.042	37.0551	76000	0.1208	19.5356	4.2369
0.055	37.5428	77000	0.1176	19.5444	4.2645
0.0534	38.0302	78000	0.1111	19.3594	4.1974
0.049	38.5179	79000	0.1184	19.4034	4.2148
0.0419	39.0054	80000	0.1148	19.3374	4.2171
0.0388	39.4931	81000	0.1137	19.3021	4.1927
0.0413	39.9807	82000	0.1171	19.2713	4.1722
0.0519	40.4682	83000	0.1162	19.1391	4.1296
0.0514	40.9559	84000	0.1189	19.2713	4.1974
0.0534	41.4433	85000	0.1144	19.1611	4.1438
0.0433	41.9310	86000	0.1134	19.1391	4.1351
0.0435	42.4184	87000	0.1187	19.1259	4.1248
0.0386	42.9061	88000	0.1106	19.1303	4.1035
0.0459	43.3936	89000	0.1152	19.1347	4.1264
0.0476	43.8812	90000	0.1179	19.0157	4.0980
0.0397	44.3687	91000	0.1197	18.9629	4.0917
0.0521	44.8564	92000	0.1174	19.0818	4.0925
0.0491	45.3438	93000	0.1153	18.9100	4.0530
0.052	45.8315	94000	0.1176	18.9585	4.0712
0.0303	46.3189	95000	0.1162	18.8703	4.0475
0.0431	46.8066	96000	0.1154	18.9364	4.0333
0.0444	47.2941	97000	0.1164	18.9629	4.0436
0.0373	47.7818	98000	0.1196	18.8659	4.0514
0.0387	48.2692	99000	0.1166	18.8307	4.0436
0.0413	48.7569	100000	0.1159	18.9496	4.0491
0.0445	49.2443	101000	0.1141	18.8836	4.0467
0.0375	49.7320	102000	0.1168	19.0906	4.0917
0.032	50.2195	103000	0.1158	18.8087	4.0372
0.0361	50.7071	104000	0.1170	18.8131	4.0025
0.0305	51.1946	105000	0.1166	18.8395	4.0222
0.0483	51.6823	106000	0.1122	18.8659	4.0080
0.028	52.1697	107000	0.1099	18.6941	3.9804
0.0305	52.6574	108000	0.1114	18.6677	3.9615
0.0367	53.1448	109000	0.1113	18.5663	3.9615
0.0249	53.6325	110000	0.1127	18.5884	3.9615
0.0316	54.1200	111000	0.1164	18.5487	3.9536
0.0357	54.6077	112000	0.1135	18.5707	3.9473
0.0358	55.0951	113000	0.1138	18.4782	3.9378
0.0353	55.5828	114000	0.1137	18.4782	3.9370
0.0237	56.0702	115000	0.1143	18.5487	3.9339
0.0273	56.5579	116000	0.1121	18.5399	3.9244
0.0312	57.0454	117000	0.1135	18.4826	3.9165
0.0317	57.5330	118000	0.1125	18.5002	3.9220
0.0286	58.0205	119000	0.1122	18.4253	3.9055
0.0312	58.5082	120000	0.1133	18.4386	3.9110
0.0214	58.9959	121000	0.1136	18.4694	3.9102
0.0292	59.4833	122000	0.1123	18.5179	3.9142
0.026	59.9710	123000	0.1124	18.4782	3.9149

Framework versions

• Transformers 4.48.1
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0