wavlm-common_voice-ur

This model is a fine-tuned version of microsoft/wavlm-large on the MOZILLA-FOUNDATION/COMMON_VOICE_11_0 - UR dataset. It achieves the following results on the evaluation set:

• Loss: inf
• Wer: 0.3781

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: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• total_train_batch_size: 8
• total_eval_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 15.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.9073	0.11	100	inf	1.0
3.3187	0.22	200	inf	1.0
2.9683	0.32	300	inf	0.9991
2.454	0.43	400	inf	0.9915
1.1169	0.54	500	inf	0.7906
1.5943	0.65	600	inf	0.7260
0.9991	0.75	700	inf	0.7305
1.0608	0.86	800	inf	0.6655
1.4739	0.97	900	inf	0.6120
0.8682	1.08	1000	inf	0.6087
0.8025	1.18	1100	inf	0.5991
0.8468	1.29	1200	inf	0.5605
0.5896	1.4	1300	inf	0.5550
0.6304	1.51	1400	inf	0.5441
0.6533	1.61	1500	inf	0.5297
0.7636	1.72	1600	inf	0.5210
0.5155	1.83	1700	inf	0.5331
0.6266	1.94	1800	inf	0.5182
0.4286	2.05	1900	inf	0.4956
0.527	2.15	2000	inf	0.4935
0.4919	2.26	2100	inf	0.4933
0.3977	2.37	2200	inf	0.5015
0.5349	2.48	2300	inf	0.4942
0.5066	2.58	2400	inf	0.4684
0.6734	2.69	2500	inf	0.4870
0.5411	2.8	2600	inf	0.4919
0.3451	2.91	2700	inf	0.4607
0.3913	3.01	2800	inf	0.4558
0.3046	3.12	2900	inf	0.4685
0.2954	3.23	3000	inf	0.4638
0.5469	3.34	3100	inf	0.4495
0.2334	3.44	3200	inf	0.4547
0.3119	3.55	3300	inf	0.4619
0.6393	3.66	3400	inf	0.4541
0.4133	3.77	3500	inf	0.4456
0.4946	3.88	3600	inf	0.4369
0.3484	3.98	3700	inf	0.4335
0.3996	4.09	3800	inf	0.4717
0.2754	4.2	3900	inf	0.4414
0.3141	4.31	4000	inf	0.4390
0.2231	4.41	4100	inf	0.4353
0.2673	4.52	4200	inf	0.4410
0.2911	4.63	4300	inf	0.4337
0.3643	4.74	4400	inf	0.4362
0.2706	4.84	4500	inf	0.4359
0.2464	4.95	4600	inf	0.4249
0.1453	5.06	4700	inf	0.4293
0.2619	5.17	4800	inf	0.4201
0.1888	5.27	4900	inf	0.4222
0.2571	5.38	5000	inf	0.4333
0.1653	5.49	5100	inf	0.4192
0.2102	5.6	5200	inf	0.4232
0.1456	5.71	5300	inf	0.4198
0.3314	5.81	5400	inf	0.4169
0.1703	5.92	5500	inf	0.4118
0.1546	6.03	5600	inf	0.4147
0.2065	6.14	5700	inf	0.4291
0.1792	6.24	5800	inf	0.4175
0.2433	6.35	5900	inf	0.4157
0.352	6.46	6000	inf	0.4083
0.2406	6.57	6100	inf	0.4341
0.2397	6.67	6200	inf	0.4185
0.2145	6.78	6300	inf	0.4147
0.1733	6.89	6400	inf	0.4150
0.1867	7.0	6500	inf	0.4154
0.612	7.1	6600	inf	0.4159
0.1413	7.21	6700	inf	0.4162
0.2074	7.32	6800	inf	0.4146
0.1362	7.43	6900	inf	0.4087
0.2971	7.53	7000	inf	0.4061
0.1443	7.64	7100	inf	0.4132
0.3066	7.75	7200	inf	0.4059
0.2163	7.86	7300	inf	0.4026
0.1251	7.97	7400	inf	0.4022
0.154	8.07	7500	inf	0.3980
0.1809	8.18	7600	inf	0.4030
0.0985	8.29	7700	inf	0.3992
0.1672	8.4	7800	inf	0.4049
0.1508	8.5	7900	inf	0.3985
0.1893	8.61	8000	inf	0.3999
0.1045	8.72	8100	inf	0.4014
0.2569	8.83	8200	inf	0.3976
0.2654	8.93	8300	inf	0.4021
0.0641	9.04	8400	inf	0.3964
0.1145	9.15	8500	inf	0.3995
0.1808	9.26	8600	inf	0.3960
0.0766	9.36	8700	inf	0.3938
0.1537	9.47	8800	inf	0.3909
0.2864	9.58	8900	inf	0.4028
0.1372	9.69	9000	inf	0.3970
0.06	9.8	9100	inf	0.3911
0.0831	9.9	9200	inf	0.3954
0.1469	10.01	9300	inf	0.3952
0.0683	10.12	9400	inf	0.3899
0.0694	10.23	9500	inf	0.3918
0.0919	10.33	9600	inf	0.3895
0.1842	10.44	9700	inf	0.3945
0.0581	10.55	9800	inf	0.3979
0.1397	10.66	9900	inf	0.3911
0.0657	10.76	10000	inf	0.3886
0.1316	10.87	10100	inf	0.3877
0.1434	10.98	10200	inf	0.3858
0.05	11.09	10300	inf	0.3842
0.0565	11.19	10400	inf	0.3873
0.1696	11.3	10500	inf	0.3873
0.0819	11.41	10600	inf	0.3901
0.0631	11.52	10700	inf	0.3927
0.1276	11.63	10800	inf	0.3868
0.1002	11.73	10900	inf	0.3848
0.081	11.84	11000	inf	0.3873
0.1745	11.95	11100	inf	0.3895
0.097	12.06	11200	inf	0.4021
0.0875	12.16	11300	inf	0.3876
0.027	12.27	11400	inf	0.3873
0.0859	12.38	11500	inf	0.3863
0.1192	12.49	11600	inf	0.3799
0.1055	12.59	11700	inf	0.3795
0.0603	12.7	11800	inf	0.3785
0.111	12.81	11900	inf	0.3783
0.0313	12.92	12000	inf	0.3800
0.0241	13.02	12100	inf	0.3796
0.1072	13.13	12200	inf	0.3803
0.1758	13.24	12300	inf	0.3809
0.1334	13.35	12400	inf	0.3794
0.1372	13.46	12500	inf	0.3798
0.1919	13.56	12600	inf	0.3791
0.1753	13.67	12700	inf	0.3781
0.294	13.78	12800	inf	0.3788
0.3132	13.89	12900	inf	0.3786
0.0486	13.99	13000	inf	0.3778
0.1199	14.1	13100	inf	0.3777
0.0381	14.21	13200	inf	0.3808
0.0875	14.32	13300	inf	0.3795
0.0122	14.42	13400	inf	0.3797
0.1417	14.53	13500	inf	0.3780
0.1754	14.64	13600	inf	0.3788
0.0426	14.75	13700	inf	0.3780
0.0309	14.85	13800	inf	0.3787
0.1447	14.96	13900	inf	0.3796

Framework versions

• Transformers 4.27.0.dev0
• Pytorch 1.13.1
• Datasets 2.8.0
• Tokenizers 0.13.2