Modelpublic

tadgeis/xlsr-1b-ckt-staged-bible-radio-to-chuklang-source-eval-stage2

sourceHugging Faceapache-2.0updated 4mo agoView on Hugging Face

0likes11downloads

Model Card

xlsr-1b-ckt-staged-bible-radio-to-chuklang-source-eval-stage2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-1b.

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: 1e-05
—trainbatchsize: 16
—evalbatchsize: 16
—seed: 42
—optimizer: Use OptimizerNames.ADAMWTORCHFUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
—lrschedulertype: linear
—lrschedulerwarmup_steps: 25
—num_epochs: 30
—mixedprecisiontraining: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.27	1.1905	50	1.3918	0.9984	0.4481
0.8981	2.3810	100	1.6118	1.0	0.5281
0.7267	3.5714	150	1.3004	0.9825	0.3741
0.6363	4.7619	200	1.4469	0.9921	0.4690
0.5127	5.9524	250	1.5113	0.9921	0.4688
0.4366	7.1429	300	1.4040	0.9857	0.4185
0.4093	8.3333	350	1.3967	0.9778	0.4013
0.3749	9.5238	400	1.3272	0.9698	0.3684
0.341	10.7143	450	1.3528	0.9778	0.3724
0.2958	11.9048	500	1.3849	0.9762	0.3742
0.2908	13.0952	550	1.4275	0.9778	0.3823
0.2839	14.2857	600	1.3558	0.9762	0.3607
0.2429	15.4762	650	1.4134	0.9698	0.3693
0.2405	16.6667	700	1.5297	0.9810	0.3905
0.2337	17.8571	750	1.4542	0.9778	0.3735
0.2555	19.0476	800	1.5012	0.9746	0.3741
0.2141	20.2381	850	1.5030	0.9778	0.3786

Framework versions

—Transformers 4.57.1
—Pytorch 2.11.0+cu130
—Datasets 4.8.5
—Tokenizers 0.22.2