Modelpublic

macarious/torgo_xlsr_finetune_M03_old

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes4downloads

Model Card

torgoxlsrfinetune_M03

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: 1.6993
—Wer: 0.3247

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: 4
—evalbatchsize: 4
—seed: 42
—gradientaccumulationsteps: 2
—totaltrainbatch_size: 8
—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
3.5549	0.94	1000	3.3500	1.0
1.6884	1.89	2000	1.7291	0.8173
0.8237	2.83	3000	1.4608	0.6568
0.6483	3.77	4000	1.1613	0.4954
0.5299	4.72	5000	1.5725	0.4824
0.4091	5.66	6000	1.4437	0.4685
0.3694	6.6	7000	1.4507	0.4397
0.3515	7.55	8000	1.4397	0.4230
0.2878	8.49	9000	1.4089	0.3980
0.2698	9.43	10000	1.4696	0.3776
0.2621	10.38	11000	1.6717	0.3905
0.2465	11.32	12000	1.6235	0.3544
0.221	12.26	13000	1.4009	0.3479
0.2169	13.21	14000	1.6981	0.3636
0.187	14.15	15000	1.4411	0.3302
0.1918	15.09	16000	1.6418	0.3377
0.1876	16.04	17000	1.4613	0.3340
0.1641	16.98	18000	1.6040	0.3312
0.1546	17.92	19000	1.6096	0.3293
0.1522	18.87	20000	1.7270	0.3265
0.1332	19.81	21000	1.6993	0.3247

Framework versions

—Transformers 4.26.1
—Pytorch 2.1.2
—Datasets 2.16.1
—Tokenizers 0.13.3