Modelpublic

scasutt/wav2vec2-large-xlsr-53_toy_train_data_masked_audio

sourceHugging Faceapache-2.0updated 4y agoView on Hugging Face

0likes3downloads

Model Card

wav2vec2-large-xlsr-53toytraindatamasked_audio

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.6445
—Wer: 0.4938

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: 8
—evalbatchsize: 8
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 16
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 1000
—num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.3761	1.05	250	3.4022	0.9954
3.0858	2.1	500	3.4684	0.9954
2.6302	3.15	750	1.7989	0.9865
1.1292	4.2	1000	0.8558	0.7355
0.8371	5.25	1250	0.7319	0.6621
0.5992	6.3	1500	0.6848	0.6147
0.5189	7.35	1750	0.6522	0.5742
0.454	8.4	2000	0.6601	0.5531
0.3896	9.45	2250	0.6138	0.5439
0.3678	10.5	2500	0.6436	0.5320
0.3232	11.55	2750	0.5920	0.5174
0.2926	12.6	3000	0.6615	0.5107
0.3041	13.65	3250	0.6311	0.5015
0.2882	14.7	3500	0.6182	0.5004
0.2868	15.75	3750	0.6266	0.4943
0.2508	16.81	4000	0.6587	0.4965
0.2563	17.86	4250	0.6634	0.4939
0.2213	18.91	4500	0.6441	0.4925
0.2255	19.96	4750	0.6445	0.4938

Framework versions

—Transformers 4.17.0
—Pytorch 1.11.0+cu102
—Datasets 2.0.0
—Tokenizers 0.11.6