whisper-small-yue-fold-2

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

• Loss: 0.4821
• Wer: 94.3203

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
• train_batch_size: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.7204	0.38	50	2.3209	128.8641
0.7581	0.75	100	0.4446	191.2477
0.2924	1.13	150	0.3421	127.1881
0.2626	1.5	200	0.3177	109.4041
0.2592	1.88	250	0.2957	93.0168
0.1538	2.26	300	0.2956	90.5028
0.151	2.63	350	0.2930	72.6257
0.1577	3.01	400	0.2837	69.7393
0.0751	3.38	450	0.2959	74.7672
0.0747	3.76	500	0.2924	71.6946
0.0527	4.14	550	0.3009	70.8566
0.0465	4.51	600	0.3071	69.7393
0.047	4.89	650	0.2967	70.7635
0.0181	5.26	700	0.3134	67.1322
0.0227	5.64	750	0.3190	73.3706
0.0254	6.02	800	0.3155	84.3575
0.0103	6.39	850	0.3334	71.4153
0.0124	6.77	900	0.3390	81.8436
0.0085	7.14	950	0.3322	75.1397
0.0094	7.52	1000	0.3422	72.3464
0.0099	7.89	1050	0.3450	74.4879
0.0044	8.27	1100	0.3585	72.2533
0.0054	8.65	1150	0.3760	74.5810
0.0066	9.02	1200	0.3639	72.4395
0.0058	9.4	1250	0.3664	72.2533
0.0035	9.77	1300	0.3702	75.9777
0.0068	10.15	1350	0.3756	72.2533
0.0037	10.53	1400	0.3746	80.8194
0.0046	10.9	1450	0.3866	72.8119
0.0021	11.28	1500	0.4072	70.3911
0.0018	11.65	1550	0.4021	72.9050
0.003	12.03	1600	0.4056	76.3501
0.002	12.41	1650	0.4037	71.7877
0.0037	12.78	1700	0.4119	71.3222
0.0004	13.16	1750	0.4214	71.4153
0.0017	13.53	1800	0.4159	71.0428
0.0013	13.91	1850	0.4408	70.7635
0.0005	14.29	1900	0.4486	67.3184
0.0002	14.66	1950	0.4426	69.5531
0.0004	15.04	2000	0.4597	69.6462
0.0003	15.41	2050	0.4699	71.5084
0.0001	15.79	2100	0.4753	72.9981
0.0001	16.17	2150	0.4797	91.1546
0.0004	16.54	2200	0.4821	94.3203

Framework versions

• Transformers 4.37.0.dev0
• Pytorch 1.12.1
• Datasets 2.15.0
• Tokenizers 0.15.0