wav2vec2-xlsr-1b-mecita-portuguese-all-text-protecao_aos_pandas

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: 0.1772
• Wer: 0.1114
• Cer: 0.0303

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
13.7229	0.93	7	4.8592	1.0	0.9996
13.7229	2.0	15	3.0023	1.0	1.0
13.7229	2.93	22	2.9290	1.0	1.0
13.7229	4.0	30	2.9842	1.0	1.0
13.7229	4.93	37	2.8453	1.0	1.0
13.7229	6.0	45	2.8120	1.0	1.0
13.7229	6.93	52	2.8162	1.0	1.0
13.7229	8.0	60	2.7843	1.0	1.0
13.7229	8.93	67	2.7823	1.0	1.0
13.7229	10.0	75	2.7434	1.0	1.0
13.7229	10.93	82	2.6364	1.0	1.0
13.7229	12.0	90	2.3797	0.9876	0.9861
13.7229	12.93	97	1.9516	0.9950	0.9771
3.3197	14.0	105	1.5396	1.0	0.7474
3.3197	14.93	112	1.1038	0.9950	0.4273
3.3197	16.0	120	0.6536	0.6733	0.1691
3.3197	16.93	127	0.4087	0.3218	0.0729
3.3197	18.0	135	0.3119	0.2252	0.0561
3.3197	18.93	142	0.2720	0.1757	0.0479
3.3197	20.0	150	0.2405	0.1584	0.0413
3.3197	20.93	157	0.2365	0.1584	0.0409
3.3197	22.0	165	0.2281	0.1510	0.0397
3.3197	22.93	172	0.1989	0.1361	0.0360
3.3197	24.0	180	0.2051	0.1287	0.0360
3.3197	24.93	187	0.2265	0.1287	0.0356
3.3197	26.0	195	0.2203	0.1287	0.0377
0.5589	26.93	202	0.2181	0.1213	0.0340
0.5589	28.0	210	0.2006	0.1238	0.0336
0.5589	28.93	217	0.1860	0.1213	0.0332
0.5589	30.0	225	0.1772	0.1114	0.0303
0.5589	30.93	232	0.1914	0.1238	0.0323
0.5589	32.0	240	0.1997	0.1238	0.0323
0.5589	32.93	247	0.1947	0.1262	0.0340
0.5589	34.0	255	0.2056	0.1213	0.0327
0.5589	34.93	262	0.1985	0.1213	0.0332
0.5589	36.0	270	0.2016	0.1213	0.0327
0.5589	36.93	277	0.1941	0.1139	0.0311
0.5589	38.0	285	0.1824	0.1238	0.0319
0.5589	38.93	292	0.1822	0.1089	0.0295
0.1503	40.0	300	0.1969	0.1163	0.0311
0.1503	40.93	307	0.1996	0.1163	0.0295
0.1503	42.0	315	0.1880	0.1089	0.0295
0.1503	42.93	322	0.2017	0.1312	0.0344
0.1503	44.0	330	0.1914	0.1163	0.0327
0.1503	44.93	337	0.1935	0.1163	0.0332
0.1503	46.0	345	0.1967	0.1139	0.0319
0.1503	46.93	352	0.1913	0.1064	0.0299
0.1503	48.0	360	0.1994	0.1114	0.0303
0.1503	48.93	367	0.1883	0.1089	0.0291
0.1503	50.0	375	0.1881	0.1139	0.0303

Framework versions

• Transformers 4.28.0
• Pytorch 2.2.1+cu121
• Datasets 2.17.0
• Tokenizers 0.13.3