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

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: 0.2362
• Wer: 0.0807
• Cer: 0.0310

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
15.3961	0.97	14	3.1122	1.0	1.0
15.3961	2.0	29	2.7786	1.0	1.0
15.3961	2.97	43	1.3300	0.7125	0.2734
15.3961	4.0	58	0.4810	0.2523	0.0738
15.3961	4.97	72	0.3453	0.15	0.0503
15.3961	6.0	87	0.2707	0.1114	0.0387
1.9985	6.97	101	0.2410	0.0966	0.0339
1.9985	8.0	116	0.2400	0.0864	0.0318
1.9985	8.97	130	0.2406	0.0739	0.0300
1.9985	10.0	145	0.2534	0.075	0.0306
1.9985	10.97	159	0.2395	0.0807	0.0314
1.9985	12.0	174	0.2366	0.0727	0.0296
1.9985	12.97	188	0.2465	0.0670	0.0282
0.2275	14.0	203	0.2459	0.0716	0.0292
0.2275	14.97	217	0.2362	0.0807	0.0310
0.2275	16.0	232	0.2571	0.0682	0.0292
0.2275	16.97	246	0.2488	0.0670	0.0288
0.2275	18.0	261	0.2506	0.0636	0.0284
0.2275	18.97	275	0.2510	0.0693	0.0288
0.2275	20.0	290	0.2517	0.0682	0.0280
0.1504	20.97	304	0.2621	0.0739	0.0302
0.1504	22.0	319	0.2628	0.075	0.0298
0.1504	22.97	333	0.2672	0.0773	0.0310
0.1504	24.0	348	0.2605	0.075	0.0310
0.1504	24.97	362	0.2644	0.0705	0.0298
0.1504	26.0	377	0.2671	0.0739	0.0308
0.1504	26.97	391	0.2637	0.0705	0.0298
0.1148	28.0	406	0.2730	0.0693	0.0300
0.1148	28.97	420	0.2668	0.0693	0.0296
0.1148	30.0	435	0.2701	0.0693	0.0302
0.1148	30.97	449	0.2696	0.0739	0.0302
0.1148	32.0	464	0.2697	0.0670	0.0284
0.1148	32.97	478	0.2671	0.0682	0.0292
0.1148	34.0	493	0.2667	0.0648	0.0292
0.0937	34.97	507	0.2747	0.0705	0.0296

Framework versions

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