wav2vec2-demo-M01

This model is a fine-tuned version of yip-i/uaspeech-pretrained on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.7099
• Wer: 1.4021

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
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.3895	0.9	500	2.9817	1.0007
3.0164	1.8	1000	2.9513	1.2954
3.0307	2.7	1500	2.8709	1.3286
3.1314	3.6	2000	2.8754	1.0
3.0395	4.5	2500	2.9289	1.0
3.2647	5.41	3000	2.8134	1.0014
2.9821	6.31	3500	2.8370	1.3901
2.9262	7.21	4000	2.8731	1.3809
2.9982	8.11	4500	4.4794	1.3958
3.0807	9.01	5000	2.8268	1.3951
2.8873	9.91	5500	2.8014	1.5336
2.8755	10.81	6000	2.8010	1.3873
3.2618	11.71	6500	3.1033	1.3463
3.0063	12.61	7000	2.7906	1.3753
2.8481	13.51	7500	2.7874	1.3837
2.876	14.41	8000	2.8239	1.0636
2.8966	15.32	8500	2.7753	1.3915
2.8839	16.22	9000	2.7874	1.3223
2.8351	17.12	9500	2.7755	1.3915
2.8185	18.02	10000	2.7600	1.3908
2.8193	18.92	10500	2.7542	1.3915
2.8023	19.82	11000	2.7528	1.3915
2.7934	20.72	11500	2.7406	1.3915
2.8043	21.62	12000	2.7419	1.3915
2.7941	22.52	12500	2.7407	1.3915
2.7854	23.42	13000	2.7277	1.3915
2.7924	24.32	13500	2.7279	1.3915
2.7644	25.23	14000	2.7217	1.3915
2.7703	26.13	14500	2.7273	1.5032
2.7821	27.03	15000	2.7265	1.3915
2.7632	27.93	15500	2.7154	1.3915
2.749	28.83	16000	2.7125	1.3958
2.7515	29.73	16500	2.7099	1.4021

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.1+cu113
• Datasets 1.18.3
• Tokenizers 0.13.2