xls-r-300m-model-2

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: 0.0100
• Wer: 0.1630
• Cer: 0.0543

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: 0.0003
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
27.3091	1.0	47	12.5816	1.0	1.0
8.1185	2.0	94	4.0610	1.0	1.0
3.5739	3.0	141	3.1872	1.0	1.0
3.1895	4.0	188	3.1399	1.0	1.0
3.1433	5.0	235	3.0083	1.0	1.0
3.01	6.0	282	2.8420	1.0	1.0
2.7873	7.0	329	2.7332	1.0	0.9802
2.6089	8.0	376	2.4772	1.0	0.8920
2.1861	9.0	423	1.9047	1.0	0.7375
1.8622	10.0	470	1.3076	1.0	0.6528
1.2815	11.0	517	0.9163	1.0	0.4975
1.0227	12.0	564	0.7301	0.9655	0.4100
0.7393	13.0	611	0.5188	0.8025	0.4622
0.5328	14.0	658	0.2949	0.6301	0.2661
0.3925	15.0	705	0.2821	0.4389	0.1348
0.3234	16.0	752	0.2217	0.3668	0.1066
0.3948	17.0	799	0.1686	0.2978	0.0995
0.2397	18.0	846	0.1839	0.2633	0.0910
0.1977	19.0	893	0.0884	0.2257	0.0593
0.2011	20.0	940	0.0583	0.1034	0.0247
0.1835	21.0	987	0.1113	0.1755	0.0487
0.132	22.0	1034	0.0680	0.0878	0.0219
0.1707	23.0	1081	0.0788	0.1755	0.0452
0.1541	24.0	1128	0.0246	0.0470	0.0106
0.085	25.0	1175	0.0224	0.0690	0.0155
0.2138	26.0	1222	0.0681	0.0878	0.0226
0.1498	27.0	1269	0.0211	0.0752	0.0183
0.1187	28.0	1316	0.0534	0.0596	0.0155
0.1166	29.0	1363	0.0388	0.0690	0.0176
0.1027	30.0	1410	0.0291	0.0502	0.0127
0.0588	31.0	1457	0.0534	0.1254	0.0311
0.0659	32.0	1504	0.0383	0.1317	0.0332
0.0736	33.0	1551	0.0396	0.1411	0.0339
0.0682	34.0	1598	0.0720	0.2539	0.0677
0.0793	35.0	1645	0.1239	0.2476	0.0882
0.0464	36.0	1692	0.0637	0.2571	0.1080
0.0453	37.0	1739	0.0229	0.2257	0.0776
0.0425	38.0	1786	0.0178	0.1787	0.0487
0.0792	39.0	1833	0.0739	0.1724	0.0593
0.063	40.0	1880	0.0274	0.1442	0.0473
0.0475	41.0	1927	0.0201	0.0940	0.0282
0.0471	42.0	1974	0.0155	0.0846	0.0219
0.0532	43.0	2021	0.1041	0.1411	0.0466
0.0485	44.0	2068	0.1166	0.1160	0.0318
0.0333	45.0	2115	0.0133	0.1317	0.0311
0.0314	46.0	2162	0.0115	0.1411	0.0367
0.0328	47.0	2209	0.0245	0.1599	0.0423
0.0315	48.0	2256	0.0172	0.1724	0.0395
0.0337	49.0	2303	0.0126	0.1034	0.0247
0.0187	50.0	2350	0.0109	0.1066	0.0240
0.0209	51.0	2397	0.0118	0.1536	0.0430
0.0158	52.0	2444	0.0122	0.1599	0.0522
0.028	53.0	2491	0.0092	0.1630	0.0508
0.024	54.0	2538	0.0091	0.1473	0.0459
0.0191	55.0	2585	0.0095	0.1473	0.0473
0.0155	56.0	2632	0.0100	0.1473	0.0466
0.0219	57.0	2679	0.0098	0.1567	0.0515
0.0156	58.0	2726	0.0099	0.1661	0.0543
0.0149	59.0	2773	0.0100	0.1693	0.0558
0.0169	60.0	2820	0.0100	0.1630	0.0543

Framework versions

• Transformers 4.28.0
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.13.3