wav2vec2-xlsr-1b-mecita-portuguese-all-text-protecao_aos_pandas-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: 0.2197
• Wer: 0.0981
• Cer: 0.0334

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
13.2168	0.98	21	2.9122	1.0	1.0
13.2168	2.0	43	2.9751	1.0	1.0
13.2168	2.98	64	2.8292	1.0	1.0
13.2168	4.0	86	2.5873	0.9992	0.9999
3.3173	4.98	107	1.0785	0.8941	0.2358
3.3173	6.0	129	0.3222	0.2305	0.0611
3.3173	6.98	150	0.2691	0.1363	0.0425
3.3173	8.0	172	0.2318	0.1168	0.0373
3.3173	8.98	193	0.2221	0.0966	0.0339
0.5524	10.0	215	0.2299	0.1028	0.0349
0.5524	10.98	236	0.2225	0.0911	0.0322
0.5524	12.0	258	0.2197	0.0981	0.0334
0.5524	12.98	279	0.2268	0.0919	0.0323
0.2169	14.0	301	0.2250	0.0966	0.0330
0.2169	14.98	322	0.2343	0.0950	0.0337
0.2169	16.0	344	0.2350	0.0942	0.0329
0.2169	16.98	365	0.2256	0.0919	0.0319
0.2169	18.0	387	0.2336	0.0802	0.0308
0.1634	18.98	408	0.2233	0.0826	0.0306
0.1634	20.0	430	0.2344	0.0826	0.0306
0.1634	20.98	451	0.2270	0.0818	0.0301
0.1634	22.0	473	0.2260	0.0857	0.0305
0.1634	22.98	494	0.2460	0.0841	0.0305
0.1322	24.0	516	0.2343	0.0748	0.0292
0.1322	24.98	537	0.2455	0.0794	0.0297
0.1322	26.0	559	0.2429	0.0787	0.0293
0.1322	26.98	580	0.2337	0.0810	0.0304
0.1123	28.0	602	0.2428	0.0794	0.0296
0.1123	28.98	623	0.2420	0.0755	0.0294
0.1123	30.0	645	0.2447	0.0787	0.0292
0.1123	30.98	666	0.2496	0.0763	0.0288
0.1123	32.0	688	0.2537	0.0787	0.0290

Framework versions

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