You can deploy this model and start earning money today!

alinerodrigues/
wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04

Name: wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04
Availability: InStock

transformers

pytorch

wav2vec2

automatic-speech-recognition

generated_from_trainer

apache-2.0

endpoints_compatible

Views

No views yet

wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04

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: 0.1671
Wer: 0.0812
Cer: 0.0271

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
21.5107	1.0	67	6.3931	0.9608	0.9566
6.0165	2.0	134	4.4922	0.9648	0.9784
4.0985	3.0	201	4.4454	0.9770	0.9791
4.0985	4.0	268	4.0621	0.9824	0.9753
3.471	5.0	335	4.0903	0.9797	0.9744
3.9535	6.0	402	3.1412	0.9770	0.9805
3.9535	7.0	469	3.7763	0.9743	0.9711
3.4618	8.0	536	4.0923	0.9642	0.9620
3.4186	9.0	603	4.2152	0.9753	0.9705
3.4186	10.0	670	2.6991	0.9811	0.9865
2.8947	11.0	737	0.4025	0.4569	0.0963
0.7229	12.0	804	0.2378	0.1302	0.0407
0.7229	13.0	871	0.2056	0.1106	0.0342
0.3162	14.0	938	0.1893	0.1099	0.0328
0.2601	15.0	1005	0.1732	0.1018	0.0316
0.2601	16.0	1072	0.1879	0.0933	0.0305
0.2292	17.0	1139	0.1710	0.0852	0.0283
0.2036	18.0	1206	0.1690	0.0937	0.0300
0.2036	19.0	1273	0.1691	0.0862	0.0282
0.174	20.0	1340	0.1737	0.0906	0.0288
0.1795	21.0	1407	0.1809	0.0889	0.0287
0.1795	22.0	1474	0.1689	0.0856	0.0283
0.1643	23.0	1541	0.1802	0.0845	0.0283
0.148	24.0	1608	0.1927	0.0944	0.0304
0.148	25.0	1675	0.1737	0.0839	0.0280
0.1482	26.0	1742	0.1749	0.0829	0.0275
0.1356	27.0	1809	0.1730	0.0862	0.0280
0.1356	28.0	1876	0.1936	0.0937	0.0303
0.1202	29.0	1943	0.1846	0.0896	0.0291
0.1175	30.0	2010	0.1853	0.0906	0.0290
0.1175	31.0	2077	0.1671	0.0812	0.0271
0.1086	32.0	2144	0.1884	0.0825	0.0269
0.1085	33.0	2211	0.2011	0.0869	0.0279
0.1085	34.0	2278	0.1895	0.0873	0.0282
0.0973	35.0	2345	0.1826	0.0879	0.0290
0.1219	36.0	2412	0.2073	0.0893	0.0285
0.1219	37.0	2479	0.2004	0.0852	0.0277
0.0912	38.0	2546	0.2037	0.0862	0.0284
0.0868	39.0	2613	0.1892	0.0852	0.0277
0.0868	40.0	2680	0.1854	0.0839	0.0276
0.09	41.0	2747	0.1875	0.0829	0.0276
0.0788	42.0	2814	0.1966	0.0801	0.0268
0.0788	43.0	2881	0.1965	0.0849	0.0275
0.0761	44.0	2948	0.1863	0.0815	0.0273
0.0759	45.0	3015	0.1922	0.0815	0.0267
0.0759	46.0	3082	0.2005	0.0805	0.0273
0.0736	47.0	3149	0.1978	0.0832	0.0276
0.0645	48.0	3216	0.2211	0.0839	0.0278
0.0645	49.0	3283	0.2108	0.0845	0.0277
0.0653	50.0	3350	0.1999	0.0822	0.0277
0.0673	51.0	3417	0.2033	0.0805	0.0276

Framework versions

Transformers 4.28.0
Pytorch 2.2.1+cu121
Datasets 2.17.0
Tokenizers 0.13.3

alinerodrigues/wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04

wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

alinerodrigues/
wav2vec2-xlsr-1b-mecita-portuguese-all-clean-04