Modelpublic

anshDelKgp/wav2vec2_finetuned_on_grid

sourceHugging Faceapache-2.0updated 3y agoView on Hugging Face

0likes5downloads

Model Card

wav2vec2finetunedon_grid

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.3047
—Wer: 0.7993
—Cer: 0.5300

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: 0.001
—trainbatchsize: 16
—evalbatchsize: 8
—seed: 42
—optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
—lrschedulertype: linear
—lrschedulerwarmup_steps: 500
—num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
6.8952	1.43	500	0.4457	0.7814	0.2382
0.8573	2.86	1000	0.3487	0.7846	0.4142
0.6617	4.29	1500	0.3471	0.7759	0.4080
0.5865	5.71	2000	0.3279	0.8131	0.5024
0.5855	7.14	2500	0.3384	0.7841	0.4356
0.554	8.57	3000	0.3230	0.7949	0.4626
0.5527	10.0	3500	0.3244	0.7706	0.4572
0.5224	11.43	4000	0.3197	0.7878	0.4891
0.5112	12.86	4500	0.3180	0.7939	0.4989
0.5037	14.29	5000	0.3234	0.7650	0.4987
0.5115	15.71	5500	0.3153	0.7712	0.4899
0.498	17.14	6000	0.3168	0.7649	0.4728
0.5031	18.57	6500	0.3138	0.7619	0.5011
0.4936	20.0	7000	0.3157	0.7575	0.5112
0.4935	21.43	7500	0.3140	0.7583	0.4985
0.495	22.86	8000	0.3080	0.7672	0.5257
0.4916	24.29	8500	0.3075	0.7782	0.5509
0.4897	25.71	9000	0.3053	0.7858	0.5241
0.4845	27.14	9500	0.3042	0.8001	0.5380
0.4894	28.57	10000	0.3038	0.8018	0.5462
0.4823	30.0	10500	0.3047	0.7993	0.5300

Framework versions

—Transformers 4.28.1
—Pytorch 2.0.0+cu118
—Datasets 2.11.0
—Tokenizers 0.13.3