w2v-bert-2.0-mdc-rutoro-asr-3.0.0

This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2288
• Wer: 0.2059
• Cer: 0.0411

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 50.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
4.5614	1.0	84	2.8697	1.0	0.9556
2.9558	2.0	168	1.1423	0.9883	0.2675
1.3943	3.0	252	0.4472	0.5657	0.1024
1.0741	4.0	336	0.3785	0.4383	0.0772
1.1728	5.0	420	0.3127	0.3658	0.0654
0.8781	6.0	504	0.2809	0.2998	0.0558
0.8013	7.0	588	0.2611	0.2898	0.0543
0.7441	8.0	672	0.2340	0.2698	0.0513
0.7162	9.0	756	0.2467	0.2529	0.0484
0.7013	10.0	840	0.2505	0.2421	0.0466
0.5552	11.0	924	0.2313	0.2270	0.0433
0.5324	12.0	1008	0.2374	0.2204	0.0435
0.4722	13.0	1092	0.2206	0.2189	0.0425
0.638	14.0	1176	0.2197	0.2070	0.0413
0.6088	15.0	1260	0.2083	0.2129	0.0416
0.4534	16.0	1344	0.2137	0.2085	0.0412
0.5218	17.0	1428	0.2246	0.2062	0.0403
0.3967	18.0	1512	0.2257	0.2021	0.0396
0.6081	19.0	1596	0.2431	0.1957	0.0398
0.239	20.0	1680	0.2029	0.2023	0.0392
0.3622	21.0	1764	0.2479	0.1985	0.0395
0.3429	22.0	1848	0.2335	0.1987	0.0403
0.5635	23.0	1932	0.2226	0.1982	0.0397
0.4587	24.0	2016	0.2288	0.2059	0.0411

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.0+cu128
• Datasets 4.4.0
• Tokenizers 0.22.1