whisper-tiny-kor-430k-hf-ep100

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: 0.4802
• Cer: 7.0810

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: 5e-05
• train_batch_size: 128
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.0276	1.0	2969	0.1394	10.5081
0.0115	2.0	5938	0.1684	9.0941
0.0082	3.0	8907	0.1932	8.4145
0.0063	4.0	11876	0.2197	7.8659
0.0052	5.0	14845	0.2516	7.7668
0.0047	6.0	17814	0.2314	8.0565
0.004	7.0	20783	0.2270	8.3521
0.0037	8.0	23752	0.2540	7.8283
0.0033	9.0	26721	0.2586	7.9098
0.0028	10.0	29690	0.2891	7.4437
0.0029	11.0	32659	0.2796	7.7976
0.0025	12.0	35628	0.2630	8.2731
0.0025	13.0	38597	0.2955	7.8518
0.0025	14.0	41566	0.2812	7.4797
0.002	15.0	44535	0.2859	7.9954
0.0023	16.0	47504	0.3172	7.2374
0.002	17.0	50473	0.3382	7.5966
0.0018	18.0	53442	0.3320	7.6383
0.0018	19.0	56411	0.3197	7.6900
0.0015	20.0	59380	0.3305	8.3678
0.0016	21.0	62349	0.3409	7.5117
0.0015	22.0	65318	0.3382	7.8556
0.0016	23.0	68287	0.3282	7.5863
0.0015	24.0	71256	0.3220	8.2449
0.0013	25.0	74225	0.3272	7.7731
0.0015	26.0	77194	0.3557	7.8019
0.0014	27.0	80163	0.3807	7.3311
0.0012	28.0	83132	0.3398	7.8117
0.0013	29.0	86101	0.3892	7.6089
0.001	30.0	89070	0.3876	7.7875
0.0011	31.0	92039	0.3942	7.3922
0.0012	32.0	95008	0.3836	8.0308
0.0011	33.0	97977	0.3745	7.9775
0.001	34.0	100946	0.3605	8.0117
0.001	35.0	103915	0.3615	7.4853
0.001	36.0	106884	0.3563	7.5916
0.0009	37.0	109853	0.3469	7.4750
0.0009	38.0	112822	0.3940	7.5919
0.0009	39.0	115791	0.3771	7.5443
0.0009	40.0	118760	0.3392	7.6593
0.0009	41.0	121729	0.3498	7.6393
0.0009	42.0	124698	0.3705	7.4474
0.0008	43.0	127667	0.3758	7.2274
0.0008	44.0	130636	0.3944	7.6919
0.0009	45.0	133605	0.3885	7.5565
0.0008	46.0	136574	0.3830	7.4628
0.0008	47.0	139543	0.3972	7.8546
0.0008	48.0	142512	0.3875	7.4916
0.0007	49.0	145481	0.3438	7.2606
0.0007	50.0	148450	0.3540	7.1581
0.0008	51.0	151419	0.3768	7.1712
0.0007	52.0	154388	0.4050	7.2286
0.0007	53.0	157357	0.3785	7.4637
0.0008	54.0	160326	0.4145	7.4800
0.0008	55.0	163295	0.4042	7.3791
0.0006	56.0	166264	0.3885	7.6994
0.0006	57.0	169233	0.4153	7.5440
0.0006	58.0	172202	0.4111	7.3408
0.0006	59.0	175171	0.4147	7.2872
0.0006	60.0	178140	0.4209	7.6270
0.0006	61.0	181109	0.4041	7.4258
0.0006	62.0	184078	0.4032	7.5324
0.0006	63.0	187047	0.4214	7.3687
0.0005	64.0	190016	0.3991	7.2750
0.0005	65.0	192985	0.3885	7.1731
0.0006	66.0	195954	0.4087	7.5063
0.0005	67.0	198923	0.3760	7.4913
0.0005	68.0	201892	0.3929	7.3314
0.0005	69.0	204861	0.4044	7.5173
0.0005	70.0	207830	0.4075	7.2712
0.0005	71.0	210799	0.4170	7.2415
0.0005	72.0	213768	0.4148	7.1142
0.0005	73.0	216737	0.4271	7.3020
0.0005	74.0	219706	0.4281	7.1863
0.0004	75.0	222675	0.4202	7.1543
0.0005	76.0	225644	0.4320	7.2910
0.0005	77.0	228613	0.4328	7.3995
0.0005	78.0	231582	0.4304	7.2255
0.0005	79.0	234551	0.4537	7.0023
0.0005	80.0	237520	0.4544	7.2048
0.0004	81.0	240489	0.4485	7.2167
0.0005	82.0	243458	0.4564	7.1794
0.0004	83.0	246427	0.4608	7.2145
0.0004	84.0	249396	0.4724	7.2098
0.0004	85.0	252365	0.4726	7.1424
0.0004	86.0	255334	0.4754	7.2832
0.0005	87.0	258303	0.4765	7.1709
0.0004	88.0	261272	0.4610	7.1358
0.0004	89.0	264241	0.4697	7.0797
0.0004	90.0	267210	0.4717	7.0913
0.0004	91.0	270179	0.4756	7.1017
0.0004	92.0	273148	0.4766	7.2089
0.0004	93.0	276117	0.4763	7.1057
0.0004	94.0	279086	0.4764	7.1101
0.0004	95.0	282055	0.4759	7.2170
0.0004	96.0	285024	0.4772	7.1104
0.0004	97.0	287993	0.4781	7.0819
0.0004	98.0	290962	0.4798	7.0897
0.0004	99.0	293931	0.4800	7.0872
0.0004	100.0	296900	0.4802	7.0810

Framework versions

• Transformers 4.36.2
• Pytorch 2.0.1+cu117
• Datasets 2.16.1
• Tokenizers 0.15.0