Modelpublic

jindaznb/torgo_tiny_finetune_M01_frozen_encoder

sourceHugging Faceapache-2.0updated 2y agoView on Hugging Face

0likes2downloads

Model Card

torgotinyfinetuneM01frozen_encoder

This model is a fine-tuned version of openai/whisper-tiny on the None dataset. It achieves the following results on the evaluation set:

—Loss: 0.2864
—Wer: 45.6706

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: 1
—seed: 42
—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
0.768	0.85	500	0.2601	25.8913
0.0914	1.7	1000	0.2569	99.8302
0.0699	2.55	1500	0.2626	39.3039
0.042	3.4	2000	0.2691	26.0611
0.0336	4.24	2500	0.2619	25.4669
0.0229	5.09	3000	0.2613	29.2020
0.0166	5.94	3500	0.2525	30.0509
0.0112	6.79	4000	0.2843	30.7301
0.0113	7.64	4500	0.2862	25.8913
0.0085	8.49	5000	0.2726	29.5416
0.0059	9.34	5500	0.2782	35.6537
0.0052	10.19	6000	0.2971	39.6435
0.0041	11.04	6500	0.2886	26.9949
0.0043	11.88	7000	0.2952	29.2869
0.0031	12.73	7500	0.2858	34.3803
0.0022	13.58	8000	0.2844	35.9083
0.0019	14.43	8500	0.2749	33.7861
0.0013	15.28	9000	0.2882	41.3413
0.0014	16.13	9500	0.2817	44.3973
0.0008	16.98	10000	0.2872	39.7284
0.0006	17.83	10500	0.2846	41.8506
0.0003	18.68	11000	0.2900	45.2462
0.0003	19.52	11500	0.2864	45.6706

Framework versions

—Transformers 4.32.0
—Pytorch 2.1.0+cu121
—Datasets 2.14.7
—Tokenizers 0.13.3