wav2vec2-large-xls-r-300m-lg-cv-130hr-v2

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: 0.4316
• Wer: 0.2045
• Cer: 0.0457

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.7427	1.0	5194	0.2769	0.3444	0.0749
0.2071	2.0	10388	0.2579	0.2897	0.0640
0.1752	3.0	15582	0.2357	0.2781	0.0608
0.1555	4.0	20776	0.2445	0.2626	0.0581
0.1437	5.0	25970	0.2393	0.2610	0.0584
0.1341	6.0	31164	0.2422	0.2572	0.0578
0.1244	7.0	36358	0.2358	0.2523	0.0562
0.1148	8.0	41552	0.2342	0.2532	0.0556
0.1077	9.0	46746	0.2475	0.2462	0.0551
0.0983	10.0	51940	0.2505	0.2439	0.0549
0.0895	11.0	57134	0.2556	0.2482	0.0556
0.0812	12.0	62328	0.2633	0.2492	0.0551
0.0727	13.0	67522	0.2726	0.2448	0.0546
0.0676	14.0	72716	0.2694	0.2464	0.0539
0.0611	15.0	77910	0.2984	0.2423	0.0533
0.0557	16.0	83104	0.2989	0.2418	0.0534
0.052	17.0	88298	0.3213	0.2402	0.0534
0.0478	18.0	93492	0.3279	0.2390	0.0532
0.0451	19.0	98686	0.3247	0.2352	0.0522
0.0422	20.0	103880	0.3452	0.2344	0.0521
0.0393	21.0	109074	0.3420	0.2384	0.0529
0.0378	22.0	114268	0.3429	0.2301	0.0509
0.036	23.0	119462	0.3442	0.2356	0.0520
0.0337	24.0	124656	0.3563	0.2276	0.0502
0.0319	25.0	129850	0.3480	0.2262	0.0499
0.0299	26.0	135044	0.3634	0.2233	0.0491
0.0281	27.0	140238	0.3593	0.2265	0.0499
0.0269	28.0	145432	0.3640	0.2242	0.0491
0.0249	29.0	150626	0.3713	0.2225	0.0491
0.0246	30.0	155820	0.3849	0.2193	0.0489
0.0228	31.0	161014	0.3869	0.2199	0.0482
0.0215	32.0	166208	0.3933	0.2182	0.0483
0.0205	33.0	171402	0.3920	0.2158	0.0471
0.0191	34.0	176596	0.3992	0.2166	0.0479
0.0183	35.0	181790	0.3969	0.2127	0.0467
0.0176	36.0	186984	0.3998	0.2138	0.0472
0.0168	37.0	192178	0.4068	0.2107	0.0464
0.016	38.0	197372	0.4216	0.2113	0.0477
0.0154	39.0	202566	0.4102	0.2102	0.0469
0.0149	40.0	207760	0.4267	0.2087	0.0465
0.0142	41.0	212954	0.4248	0.2097	0.0469
0.0136	42.0	218148	0.4254	0.2074	0.0467
0.0133	43.0	223342	0.4304	0.2074	0.0465
0.0131	44.0	228536	0.4314	0.2063	0.0462
0.0126	45.0	233730	0.4267	0.2047	0.0458
0.0122	46.0	238924	0.4291	0.2049	0.0457
0.012	47.0	244118	0.4305	0.2048	0.0458
0.0122	48.0	249312	0.4321	0.2046	0.0457
0.0121	49.0	254506	0.4316	0.2045	0.0457
0.0122	50.0	259700	0.4318	0.2045	0.0457

Framework versions

• Transformers 4.42.3
• Pytorch 2.2.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1