wav2vec2-xls-r-1b-ft-btb-ccv-cy

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b on the DEWIBRYNJONES/BANC-TRAWSGRIFIADAU-BANGOR-CLEAN-WITH-CCV - DEFAULT dataset.

It achieves the following results on the test set:

• Wer: 0.4416104502872399

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: 200000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.5324	0.0383	1000	inf	0.9162
1.3344	0.0765	2000	inf	0.9002
1.2517	0.1148	3000	inf	0.8682
1.2045	0.1531	4000	inf	0.8538
1.1725	0.1913	5000	inf	0.8389
1.1243	0.2296	6000	inf	0.8252
1.1034	0.2679	7000	inf	0.8147
1.0795	0.3061	8000	inf	0.8072
1.057	0.3444	9000	inf	0.7895
1.0169	0.3827	10000	inf	0.7693
1.0014	0.4209	11000	inf	0.7761
0.9994	0.4592	12000	inf	0.7627
0.9603	0.4975	13000	inf	0.7659
0.947	0.5357	14000	inf	0.7499
0.9201	0.5740	15000	inf	0.7423
0.9243	0.6123	16000	inf	0.7311
0.8956	0.6505	17000	inf	0.7158
0.9164	0.6888	18000	inf	0.7104
0.8813	0.7271	19000	inf	0.7138
0.8627	0.7653	20000	inf	0.6857
0.8618	0.8036	21000	inf	0.6947
0.8635	0.8419	22000	inf	0.6986
0.8396	0.8801	23000	inf	0.6784
0.8303	0.9184	24000	inf	0.6918
0.8217	0.9567	25000	inf	0.6671
0.8161	0.9949	26000	inf	0.6990
0.7745	1.0332	27000	inf	0.6596
0.7775	1.0715	28000	inf	0.6444
0.7677	1.1098	29000	inf	0.6411
0.7651	1.1480	30000	inf	0.6360
0.7495	1.1863	31000	inf	0.6383
0.7584	1.2246	32000	inf	0.6299
0.7467	1.2628	33000	inf	0.6328
0.7321	1.3011	34000	inf	0.6344
0.7458	1.3394	35000	inf	0.6259
0.7224	1.3776	36000	inf	0.6136
0.7149	1.4159	37000	inf	0.6039
0.7073	1.4542	38000	inf	0.6153
0.7121	1.4924	39000	inf	0.6036
0.7123	1.5307	40000	inf	0.6111
0.7059	1.5690	41000	inf	0.5991
0.6765	1.6072	42000	inf	0.5985
0.6928	1.6455	43000	inf	0.5948
0.6849	1.6838	44000	inf	0.6151
0.694	1.7220	45000	inf	0.5855
0.6786	1.7603	46000	inf	0.5924
0.6703	1.7986	47000	inf	0.5924
0.6573	1.8368	48000	inf	0.5716
0.6672	1.8751	49000	inf	0.5795
0.6401	1.9134	50000	inf	0.5739
0.6621	1.9516	51000	inf	0.5727
0.644	1.9899	52000	inf	0.5558
0.5941	2.0282	53000	inf	0.5625
0.5991	2.0664	54000	inf	0.5510
0.6077	2.1047	55000	inf	0.5534
0.5878	2.1430	56000	inf	0.5430
0.5875	2.1812	57000	inf	0.5503
0.5898	2.2195	58000	inf	0.5392
0.597	2.2578	59000	inf	0.5391
0.5769	2.2960	60000	inf	0.5348
0.5669	2.3343	61000	inf	0.5410
0.577	2.3726	62000	inf	0.5350
0.5675	2.4108	63000	inf	0.5298
0.5665	2.4491	64000	inf	0.5333
0.5782	2.4874	65000	inf	0.5259
0.5748	2.5256	66000	inf	0.5226
0.5813	2.5639	67000	inf	0.5234
0.5698	2.6022	68000	inf	0.5208
0.5606	2.6404	69000	inf	0.5123
0.5654	2.6787	70000	inf	0.5160
0.5657	2.7170	71000	inf	0.5297
0.5631	2.7552	72000	inf	0.5190
0.5442	2.7935	73000	inf	0.5100
0.5489	2.8318	74000	inf	0.5116
0.541	2.8700	75000	inf	0.5017
0.5384	2.9083	76000	inf	0.5002
0.5508	2.9466	77000	inf	0.5109
0.5174	2.9848	78000	inf	0.5033
0.4844	3.0231	79000	inf	0.4903
0.4768	3.0614	80000	inf	0.5
0.4804	3.0996	81000	inf	0.4968
0.4873	3.1379	82000	inf	0.4874
0.4805	3.1762	83000	inf	0.4860
0.4763	3.2144	84000	inf	0.4849
0.483	3.2527	85000	inf	0.4902
0.473	3.2910	86000	inf	0.4808
0.4747	3.3293	87000	inf	0.4906
0.4722	3.3675	88000	inf	0.4765
0.4662	3.4058	89000	inf	0.4763
0.4683	3.4441	90000	inf	0.4785
0.4686	3.4823	91000	inf	0.4726
0.4519	3.5206	92000	inf	0.4771
0.4509	3.5589	93000	inf	0.4737
0.4679	3.5971	94000	inf	0.4712
0.4693	3.6354	95000	inf	0.4898
0.4615	3.6737	96000	inf	0.4688
0.4671	3.7119	97000	inf	0.4681
0.445	3.7502	98000	inf	0.4604
0.4559	3.7885	99000	inf	0.4598
0.4514	3.8267	100000	inf	0.4554
0.4501	3.8650	101000	inf	0.4521
0.4488	3.9033	102000	inf	0.4569
0.472	3.9415	103000	inf	0.4532
0.4603	3.9798	104000	inf	0.4536
0.3935	4.0181	105000	inf	0.4460
0.3981	4.0563	106000	inf	0.4440
0.3935	4.0946	107000	inf	0.4461
0.3936	4.1329	108000	inf	0.4410
0.3902	4.1711	109000	inf	0.4411
0.388	4.2094	110000	inf	0.4417
0.393	4.2477	111000	inf	0.4347
0.3957	4.2859	112000	inf	0.4398
0.3801	4.3242	113000	inf	0.4380
0.3836	4.3625	114000	inf	0.4327
0.383	4.4007	115000	inf	0.4385
0.38	4.4390	116000	inf	0.4374
0.4004	4.4773	117000	inf	0.4265
0.3817	4.5155	118000	inf	0.4342
0.3937	4.5538	119000	inf	0.4295
0.3891	4.5921	120000	inf	0.4257
0.3883	4.6303	121000	inf	0.4256
0.3691	4.6686	122000	inf	0.4195
0.3699	4.7069	123000	inf	0.4273
0.37	4.7451	124000	inf	0.4203
0.3691	4.7834	125000	inf	0.4201
0.3701	4.8217	126000	inf	0.4120
0.3701	4.8599	127000	inf	0.4193
0.3732	4.8982	128000	inf	0.4195
0.3755	4.9365	129000	inf	0.4196
0.3778	4.9747	130000	inf	0.4098
0.3384	5.0130	131000	inf	0.4119
0.3216	5.0513	132000	inf	0.4103
0.3149	5.0895	133000	inf	0.4123
0.3162	5.1278	134000	inf	0.4070
0.3206	5.1661	135000	inf	0.4083
0.3137	5.2043	136000	inf	0.4067
0.3202	5.2426	137000	inf	0.4083
0.3076	5.2809	138000	inf	0.4038
0.3172	5.3191	139000	inf	0.4020
0.3235	5.3574	140000	inf	0.4031
0.3019	5.3957	141000	inf	0.3973
0.3085	5.4340	142000	inf	0.4003
0.3177	5.4722	143000	inf	0.3985
0.321	5.5105	144000	inf	0.3982
0.325	5.5488	145000	inf	0.3972
0.308	5.5870	146000	inf	0.3980
0.318	5.6253	147000	inf	0.3931
0.3124	5.6636	148000	inf	0.3936
0.3039	5.7018	149000	inf	0.3911
0.3028	5.7401	150000	inf	0.3902
0.3083	5.7784	151000	inf	0.3890
0.3004	5.8166	152000	inf	0.3894
0.2987	5.8549	153000	inf	0.3823
0.3048	5.8932	154000	inf	0.3838
0.2959	5.9314	155000	inf	0.3799
0.301	5.9697	156000	inf	0.3831
0.2778	6.0080	157000	inf	0.3792
0.2549	6.0462	158000	inf	0.3790
0.2596	6.0845	159000	inf	0.3775
0.2436	6.1228	160000	inf	0.3795
0.2489	6.1610	161000	inf	0.3765
0.2516	6.1993	162000	inf	0.3767
0.2565	6.2376	163000	inf	0.3772
0.2482	6.2758	164000	inf	0.3721
0.2561	6.3141	165000	inf	0.3740
0.2497	6.3524	166000	inf	0.3729
0.2544	6.3906	167000	inf	0.3714
0.2504	6.4289	168000	inf	0.3700
0.2502	6.4672	169000	inf	0.3710
0.2493	6.5054	170000	inf	0.3683
0.2586	6.5437	171000	inf	0.3696
0.2502	6.5820	172000	inf	0.3670
0.2551	6.6202	173000	inf	0.3681
0.2556	6.6585	174000	inf	0.3651
0.2539	6.6968	175000	inf	0.3651
0.2464	6.7350	176000	inf	0.3639
0.2395	6.7733	177000	inf	0.3624
0.2434	6.8116	178000	inf	0.3637
0.2384	6.8498	179000	inf	0.3608
0.2339	6.8881	180000	inf	0.3615
0.2392	6.9264	181000	inf	0.3608
0.2348	6.9646	182000	inf	0.3615
0.2232	7.0029	183000	inf	0.3592
0.213	7.0412	184000	inf	0.3604
0.2036	7.0794	185000	inf	0.3589
0.2055	7.1177	186000	inf	0.3585
0.202	7.1560	187000	inf	0.3559
0.2004	7.1942	188000	inf	0.3552
0.2049	7.2325	189000	inf	0.3547
0.2013	7.2708	190000	inf	0.3566
0.2015	7.3090	191000	inf	0.3547
0.2043	7.3473	192000	inf	0.3536
0.2023	7.3856	193000	inf	0.3545
0.2002	7.4238	194000	inf	0.3535
0.199	7.4621	195000	inf	0.3540
0.2056	7.5004	196000	inf	0.3539
0.203	7.5386	197000	inf	0.3534
0.194	7.5769	198000	inf	0.3531
0.2076	7.6152	199000	inf	0.3529
0.1931	7.6535	200000	inf	0.3532

Framework versions

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