wav2vec2-xlsr-53-ft-btb-ccv-enc-cy

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-CLEAN-WITH-CCV - DEFAULT dataset. It achieves the following results on the evaluation set:

• Loss: 0.4095
• Wer: 0.3271

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_steps: 500
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.0194	100	3.5475	1.0
No log	0.0387	200	3.0259	1.0
No log	0.0581	300	3.0887	1.0
No log	0.0774	400	2.3822	0.9972
4.0938	0.0968	500	1.4547	0.9020
4.0938	0.1161	600	1.2603	0.8510
4.0938	0.1355	700	1.0940	0.7655
4.0938	0.1549	800	1.0705	0.7602
4.0938	0.1742	900	0.9356	0.6973
1.0597	0.1936	1000	0.9104	0.6766
1.0597	0.2129	1100	0.8879	0.6570
1.0597	0.2323	1200	0.8595	0.6612
1.0597	0.2516	1300	0.8352	0.6075
1.0597	0.2710	1400	0.7912	0.6033
0.8484	0.2904	1500	0.7862	0.6067
0.8484	0.3097	1600	0.7790	0.6009
0.8484	0.3291	1700	0.7678	0.5629
0.8484	0.3484	1800	0.7515	0.5799
0.8484	0.3678	1900	0.7424	0.5859
0.764	0.3871	2000	0.7130	0.5521
0.764	0.4065	2100	0.7114	0.5408
0.764	0.4259	2200	0.7229	0.5577
0.764	0.4452	2300	0.6773	0.5160
0.764	0.4646	2400	0.6784	0.5178
0.6868	0.4839	2500	0.6720	0.5262
0.6868	0.5033	2600	0.6804	0.5337
0.6868	0.5226	2700	0.6599	0.5024
0.6868	0.5420	2800	0.6287	0.4902
0.6868	0.5614	2900	0.6304	0.4947
0.6761	0.5807	3000	0.6258	0.4851
0.6761	0.6001	3100	0.6311	0.4990
0.6761	0.6194	3200	0.6172	0.4901
0.6761	0.6388	3300	0.6187	0.4666
0.6761	0.6581	3400	0.6045	0.4725
0.6462	0.6775	3500	0.5950	0.4717
0.6462	0.6969	3600	0.5903	0.4602
0.6462	0.7162	3700	0.5865	0.4727
0.6462	0.7356	3800	0.5820	0.4590
0.6462	0.7549	3900	0.6026	0.4830
0.6193	0.7743	4000	0.5807	0.4496
0.6193	0.7937	4100	0.5621	0.4486
0.6193	0.8130	4200	0.5730	0.4593
0.6193	0.8324	4300	0.5592	0.4374
0.6193	0.8517	4400	0.5621	0.4239
0.59	0.8711	4500	0.5458	0.4304
0.59	0.8904	4600	0.5406	0.4271
0.59	0.9098	4700	0.5269	0.4132
0.59	0.9292	4800	0.5362	0.4215
0.59	0.9485	4900	0.5226	0.4163
0.5636	0.9679	5000	0.5297	0.4148
0.5636	0.9872	5100	0.5226	0.4136
0.5636	1.0066	5200	0.5239	0.4054
0.5636	1.0259	5300	0.5383	0.4058
0.5636	1.0453	5400	0.5125	0.4067
0.4924	1.0647	5500	0.5029	0.3953
0.4924	1.0840	5600	0.5054	0.3932
0.4924	1.1034	5700	0.4969	0.3894
0.4924	1.1227	5800	0.4935	0.3851
0.4924	1.1421	5900	0.4977	0.3817
0.4602	1.1614	6000	0.4863	0.3874
0.4602	1.1808	6100	0.4906	0.3777
0.4602	1.2002	6200	0.4891	0.3764
0.4602	1.2195	6300	0.4881	0.3801
0.4602	1.2389	6400	0.4814	0.3727
0.4407	1.2582	6500	0.4714	0.3772
0.4407	1.2776	6600	0.4739	0.3706
0.4407	1.2969	6700	0.4692	0.3714
0.4407	1.3163	6800	0.4673	0.3728
0.4407	1.3357	6900	0.4610	0.3678
0.4284	1.3550	7000	0.4730	0.3653
0.4284	1.3744	7100	0.4606	0.3640
0.4284	1.3937	7200	0.4572	0.3620
0.4284	1.4131	7300	0.4575	0.3630
0.4284	1.4324	7400	0.4578	0.3590
0.4299	1.4518	7500	0.4477	0.3569
0.4299	1.4712	7600	0.4442	0.3552
0.4299	1.4905	7700	0.4420	0.3546
0.4299	1.5099	7800	0.4437	0.3483
0.4299	1.5292	7900	0.4373	0.3486
0.408	1.5486	8000	0.4336	0.3464
0.408	1.5679	8100	0.4348	0.3448
0.408	1.5873	8200	0.4276	0.3418
0.408	1.6067	8300	0.4294	0.3399
0.408	1.6260	8400	0.4272	0.3388
0.3964	1.6454	8500	0.4311	0.3409
0.3964	1.6647	8600	0.4260	0.3381
0.3964	1.6841	8700	0.4260	0.3371
0.3964	1.7034	8800	0.4260	0.3364
0.3964	1.7228	8900	0.4215	0.3351
0.3866	1.7422	9000	0.4234	0.3330
0.3866	1.7615	9100	0.4210	0.3319
0.3866	1.7809	9200	0.4156	0.3301
0.3866	1.8002	9300	0.4158	0.3303
0.3866	1.8196	9400	0.4155	0.3294
0.37	1.8389	9500	0.4137	0.3292
0.37	1.8583	9600	0.4120	0.3284
0.37	1.8777	9700	0.4109	0.3301
0.37	1.8970	9800	0.4100	0.3279
0.37	1.9164	9900	0.4095	0.3267
0.371	1.9357	10000	0.4095	0.3271

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1