wav2vec-bert-2.0-ulch-try

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 1.3563
• Wer: 0.4872
• Cer: 0.1704

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: 5e-05
• train_batch_size: 1
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 8
• 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: 500
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.4139	0.1219	180	1.7147	0.9451	0.4056
1.5054	0.2438	360	1.4105	0.8582	0.3543
1.4072	0.3656	540	1.3314	0.8832	0.3701
1.236	0.4875	720	1.1863	0.7590	0.2799
1.1715	0.6094	900	1.1086	0.7380	0.2837
1.1484	0.7313	1080	1.0823	0.6931	0.2509
1.1007	0.8532	1260	1.0204	0.7211	0.2560
1.0404	0.9750	1440	0.9578	0.7289	0.2511
0.9031	1.0968	1620	0.9397	0.6513	0.2303
0.9951	1.2187	1800	0.9252	0.6504	0.2309
0.842	1.3406	1980	0.9184	0.6139	0.2190
0.8247	1.4625	2160	0.8968	0.6143	0.2208
0.867	1.5843	2340	0.8945	0.6140	0.2192
0.8674	1.7062	2520	0.8362	0.6092	0.2097
0.818	1.8281	2700	0.8848	0.6116	0.2130
0.7734	1.9500	2880	0.8408	0.5740	0.2002
0.7266	2.0718	3060	0.8358	0.5806	0.2041
0.6954	2.1937	3240	0.8125	0.5705	0.1985
0.6441	2.3155	3420	0.8175	0.5714	0.1985
0.6673	2.4374	3600	0.8032	0.5637	0.1974
0.6947	2.5593	3780	0.8041	0.5884	0.2165
0.6856	2.6812	3960	0.7781	0.5513	0.1952
0.6401	2.8030	4140	0.7975	0.5548	0.1935
0.6342	2.9249	4320	0.7842	0.5377	0.1899
0.6434	3.0467	4500	0.7777	0.5395	0.1874
0.5545	3.1686	4680	0.7825	0.5479	0.1941
0.5798	3.2905	4860	0.7889	0.5335	0.1866
0.5142	3.4124	5040	0.7797	0.5371	0.1918
0.5413	3.5342	5220	0.7797	0.5461	0.1891
0.5282	3.6561	5400	0.7814	0.5357	0.1868
0.5308	3.7780	5580	0.7649	0.5202	0.1802
0.4949	3.8999	5760	0.7638	0.5458	0.1856
0.4853	4.0217	5940	0.7786	0.5247	0.1803
0.4026	4.1435	6120	0.7678	0.5277	0.1857
0.428	4.2654	6300	0.7772	0.5217	0.1795
0.394	4.3873	6480	0.7642	0.5118	0.1779
0.418	4.5092	6660	0.7672	0.5219	0.1809
0.4245	4.6311	6840	0.7464	0.5113	0.1811
0.4657	4.7529	7020	0.7470	0.5211	0.1859
0.4285	4.8748	7200	0.7455	0.5269	0.1858
0.416	4.9967	7380	0.7357	0.5325	0.1860
0.3282	5.1185	7560	0.8268	0.5086	0.1791
0.3192	5.2404	7740	0.8087	0.5088	0.1781
0.3262	5.3623	7920	0.7779	0.5161	0.1836
0.3152	5.4841	8100	0.8079	0.5064	0.1766
0.3622	5.6060	8280	0.8168	0.5062	0.1789
0.318	5.7279	8460	0.8088	0.4976	0.1744
0.3107	5.8498	8640	0.8074	0.5020	0.1792
0.3017	5.9716	8820	0.7807	0.5068	0.1759
0.2594	6.0934	9000	0.9053	0.5003	0.1740
0.2994	6.2153	9180	0.8920	0.5022	0.1746
0.2288	6.3372	9360	0.9006	0.4981	0.1773
0.2345	6.4591	9540	0.8932	0.4963	0.1761
0.2287	6.5810	9720	0.8502	0.4961	0.1734
0.2291	6.7028	9900	0.8334	0.5010	0.1787
0.2075	6.8247	10080	0.8773	0.4869	0.1726
0.2182	6.9466	10260	0.8423	0.4964	0.1754
0.1815	7.0684	10440	0.9376	0.4936	0.1733
0.1586	7.1903	10620	0.9601	0.5016	0.1751
0.1578	7.3121	10800	0.9650	0.4894	0.1726
0.1489	7.4340	10980	0.9943	0.4852	0.1700
0.153	7.5559	11160	1.0456	0.4832	0.1686
0.1501	7.6778	11340	1.0008	0.4857	0.1714
0.1397	7.7997	11520	0.9729	0.4878	0.1740
0.1554	7.9215	11700	0.9443	0.4826	0.1710
0.1732	8.0433	11880	1.0668	0.4854	0.1707
0.0863	8.1652	12060	1.1487	0.4849	0.1708
0.0886	8.2871	12240	1.1944	0.4866	0.1714
0.0986	8.4090	12420	1.1378	0.4818	0.1691
0.137	8.5309	12600	1.1633	0.4923	0.1716
0.0954	8.6527	12780	1.1348	0.4910	0.1713
0.101	8.7746	12960	1.0933	0.4883	0.1715
0.0833	8.8965	13140	1.1138	0.4951	0.1721
0.1005	9.0183	13320	1.1980	0.4866	0.1699
0.0476	9.1402	13500	1.3154	0.4908	0.1721
0.0529	9.2620	13680	1.3140	0.4874	0.1720
0.0532	9.3839	13860	1.3547	0.4873	0.1711
0.0499	9.5058	14040	1.3280	0.4850	0.1704
0.0538	9.6277	14220	1.3608	0.4858	0.1705
0.0529	9.7496	14400	1.3606	0.4838	0.1704
0.0504	9.8714	14580	1.3553	0.4876	0.1706
0.0411	9.9933	14760	1.3563	0.4872	0.1704

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 3.5.1
• Tokenizers 0.21.1