Modelpublic

dolphinnlp/model_wav2vec_

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes5downloads

Model Card

modelwav2vec

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.4608
—Wer: 0.3361

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: 0.0001
—trainbatchsize: 16
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 30
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.5101	2.01	500	1.6030	1.0221
0.7158	4.02	1000	0.4599	0.4587
0.2914	6.02	1500	0.4092	0.4037
0.1906	8.03	2000	0.3927	0.3865
0.1328	10.04	2500	0.4067	0.3810
0.1122	12.05	3000	0.4128	0.3645
0.0954	14.06	3500	0.4469	0.3621
0.0754	16.06	4000	0.4296	0.3639
0.0589	18.07	4500	0.4434	0.3567
0.0497	20.08	5000	0.4597	0.3535
0.0455	22.09	5500	0.4854	0.3454
0.0395	24.1	6000	0.4615	0.3435
0.0328	26.1	6500	0.4714	0.3396
0.0284	28.11	7000	0.4608	0.3361

Framework versions

—Transformers 4.36.2
—Pytorch 2.1.0+cu121
—Datasets 1.18.3
—Tokenizers 0.15.0