wav2vec2-large-xlsr-mecita-coraa-portuguese-random-all-02

This model is a fine-tuned version of jonatasgrosman/wav2vec2-xls-r-1b-portuguese on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2941
• Wer: 0.1192
• Cer: 0.0385

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: 3e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
26.6237	1.0	86	2.7490	0.9342	0.7322
5.3366	2.0	172	1.6394	0.7008	0.4202
3.393	3.0	258	1.0152	0.4928	0.2308
2.1045	4.0	344	0.6938	0.3511	0.1511
1.8687	5.0	430	0.5804	0.2975	0.1133
1.0808	6.0	516	0.4939	0.2553	0.0986
0.9453	7.0	602	0.4599	0.2317	0.0858
0.9453	8.0	688	0.4172	0.2139	0.0739
0.7718	9.0	774	0.3793	0.1807	0.0622
0.72	10.0	860	0.4070	0.1814	0.0636
0.4718	11.0	946	0.4255	0.1644	0.0570
0.5514	12.0	1032	0.3529	0.1526	0.0511
0.3473	13.0	1118	0.3458	0.1529	0.0505
0.4885	14.0	1204	0.3217	0.1444	0.0496
0.4885	15.0	1290	0.3107	0.1448	0.0480
0.3774	16.0	1376	0.4008	0.1461	0.0488
0.3974	17.0	1462	0.3857	0.1519	0.0501
0.3398	18.0	1548	0.3497	0.1309	0.0422
0.2741	19.0	1634	0.3150	0.1312	0.0426
0.2805	20.0	1720	0.3533	0.1236	0.0405
0.3292	21.0	1806	0.3227	0.1278	0.0424
0.3292	22.0	1892	0.2969	0.1295	0.0416
0.2255	23.0	1978	0.2941	0.1192	0.0385
0.2107	24.0	2064	0.3290	0.1261	0.0421
0.1922	25.0	2150	0.3492	0.1222	0.0399
0.1829	26.0	2236	0.3640	0.1173	0.0383
0.1911	27.0	2322	0.3595	0.1224	0.0396
0.1712	28.0	2408	0.3521	0.1219	0.0390
0.1712	29.0	2494	0.3313	0.1136	0.0374
0.1708	30.0	2580	0.3219	0.1207	0.0381
0.1389	31.0	2666	0.3261	0.1114	0.0361
0.1516	32.0	2752	0.3446	0.1102	0.0359
0.2601	33.0	2838	0.3505	0.1151	0.0367
0.1392	34.0	2924	0.3282	0.1131	0.0367
0.1286	35.0	3010	0.3351	0.1129	0.0359
0.1286	36.0	3096	0.3482	0.1119	0.0357
0.1497	37.0	3182	0.3762	0.1156	0.0373
0.1319	38.0	3268	0.3733	0.1149	0.0373
0.1294	39.0	3354	0.3463	0.1178	0.0372
0.1459	40.0	3440	0.3440	0.1146	0.0378
0.0998	41.0	3526	0.3467	0.1131	0.0366
0.1036	42.0	3612	0.3225	0.1127	0.0362
0.1036	43.0	3698	0.3630	0.1105	0.0351

Framework versions

• Transformers 4.28.0
• Pytorch 2.5.0+cu121
• Datasets 3.1.0
• Tokenizers 0.13.3