README.md

metadata

library_name: transformers
language:
  - lg
license: mit
base_model: facebook/w2v-bert-2.0
tags:
  - generated_from_trainer
datasets:
  - yogera
metrics:
  - wer
model-index:
  - name: wav2vec2-bert
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Yogera
          type: yogera
        metrics:
          - name: Wer
            type: wer
            value: 0.18076661374461328

wav2vec2-bert

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

• Loss: 0.3420
• Wer: 0.1808
• Cer: 0.0385

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: 8
• seed: 42
• 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
2.0247	1.0	57	0.3818	0.4570	0.0903
0.3116	2.0	114	0.2514	0.3059	0.0614
0.2069	3.0	171	0.2253	0.2704	0.0555
0.1586	4.0	228	0.2139	0.2601	0.0506
0.1293	5.0	285	0.2124	0.2200	0.0458
0.1052	6.0	342	0.2168	0.2087	0.0438
0.0864	7.0	399	0.2385	0.2110	0.0444
0.0791	8.0	456	0.2177	0.2030	0.0424
0.068	9.0	513	0.2356	0.2002	0.0422
0.0609	10.0	570	0.2482	0.2083	0.0429
0.0496	11.0	627	0.2482	0.1977	0.0438
0.0467	12.0	684	0.2556	0.1978	0.0419
0.0398	13.0	741	0.2688	0.1960	0.0409
0.0369	14.0	798	0.2580	0.1951	0.0411
0.0349	15.0	855	0.2673	0.1989	0.0426
0.0333	16.0	912	0.2926	0.1936	0.0413
0.0295	17.0	969	0.2854	0.1962	0.0408
0.0241	18.0	1026	0.2841	0.1888	0.0406
0.0221	19.0	1083	0.2928	0.1954	0.0419
0.0213	20.0	1140	0.3104	0.2041	0.0436
0.0208	21.0	1197	0.2975	0.1881	0.0416
0.0217	22.0	1254	0.2764	0.1913	0.0417
0.0193	23.0	1311	0.2933	0.1928	0.0419
0.0135	24.0	1368	0.3073	0.1859	0.0401
0.0146	25.0	1425	0.2925	0.1851	0.0401
0.0156	26.0	1482	0.3120	0.1898	0.0407
0.0137	27.0	1539	0.2996	0.1915	0.0428
0.0116	28.0	1596	0.3228	0.1859	0.0409
0.011	29.0	1653	0.3331	0.1915	0.0416
0.0116	30.0	1710	0.3209	0.1804	0.0391
0.0083	31.0	1767	0.3333	0.1862	0.0400
0.006	32.0	1824	0.3504	0.1854	0.0398
0.0075	33.0	1881	0.3189	0.1965	0.0413
0.0054	34.0	1938	0.3461	0.1837	0.0398
0.0074	35.0	1995	0.3263	0.1867	0.0402
0.0071	36.0	2052	0.3430	0.1876	0.0401
0.0068	37.0	2109	0.3381	0.1913	0.0411
0.0097	38.0	2166	0.3150	0.1937	0.0416
0.0079	39.0	2223	0.3313	0.1856	0.0405
0.0062	40.0	2280	0.3420	0.1808	0.0385

Framework versions

• Transformers 4.45.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.1
• Tokenizers 0.20.1