whisper_tn_hi

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: 1.6400
• Wer: 147.3885
• Cer: 295.3741

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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	Wer	Cer
1.6414	1.0	409	0.8809	83.9795	248.4138
0.531	2.0	818	0.6346	75.4042	280.7284
0.3538	3.0	1227	0.5810	69.8595	259.1276
0.2679	4.0	1636	0.5639	84.1488	332.292
0.2074	5.0	2045	0.5715	85.9773	264.435
0.1599	6.0	2454	0.6074	86.481	262.8894
0.1216	7.0	2863	0.6402	110.0398	285.4649
0.0903	8.0	3272	0.6736	92.7961	278.7952
0.0663	9.0	3681	0.7023	96.3896	268.8475
0.0472	10.0	4090	0.7527	104.0041	276.3805
0.0335	11.0	4499	0.7907	99.0646	274.9479
0.0235	12.0	4908	0.8320	128.1004	282.2682
0.0169	13.0	5317	0.8741	116.1305	277.659
0.0124	14.0	5726	0.9090	137.6534	290.6719
0.0094	15.0	6135	0.9492	117.0405	285.2422
0.007	16.0	6544	0.9905	122.1663	280.3801
0.0061	17.0	6953	1.0199	125.0656	277.5049
0.0051	18.0	7362	1.0383	117.3368	278.8806
0.0044	19.0	7771	1.0617	110.8736	275.7667
0.0041	20.0	8180	1.0867	142.9061	291.1584
0.0037	21.0	8589	1.1224	119.377	273.63
0.0026	22.0	8998	1.1322	158.2452	295.0852
0.0024	23.0	9407	1.1619	134.9446	283.1038
0.0022	24.0	9816	1.1677	124.5789	283.5701
0.002	25.0	10225	1.1898	125.0275	288.5318
0.0019	26.0	10634	1.1994	138.0386	288.9011
0.0023	27.0	11043	1.2216	119.7071	279.7329
0.0021	28.0	11452	1.2521	96.3388	266.6656
0.0018	29.0	11861	1.2568	148.4932	288.7504
0.0018	30.0	12270	1.2541	115.3771	283.8205
0.0021	31.0	12679	1.2291	98.8995	271.6817
0.0014	32.0	13088	1.2821	130.6654	293.2197
0.0014	33.0	13497	1.2804	121.8954	287.2249
0.0013	34.0	13906	1.2802	137.5857	293.6275
0.0015	35.0	14315	1.3010	147.5789	296.7907
0.0014	36.0	14724	1.2945	139.6766	292.4335
0.0012	37.0	15133	1.3310	144.5653	288.2045
0.0011	38.0	15542	1.3200	160.6493	306.8297
0.0009	39.0	15951	1.3394	211.9783	341.8621
0.0013	40.0	16360	1.3367	133.4166	304.8621
0.0007	41.0	16769	1.3472	154.6601	319.8702
0.0005	42.0	17178	1.3617	149.0815	301.4669
0.0009	43.0	17587	1.3570	163.2312	319.4675
0.0009	44.0	17996	1.3723	149.9915	310.0088
0.0009	45.0	18405	1.3809	133.1118	289.5232
0.0009	46.0	18814	1.3664	166.6427	308.9287
0.0008	47.0	19223	1.3894	150.0127	304.3739
0.0005	48.0	19632	1.3632	129.4929	307.7766
0.0005	49.0	20041	1.3917	143.9304	313.3529
0.0005	50.0	20450	1.4006	113.0111	295.1966
0.0007	51.0	20859	1.3966	158.3129	303.4328
0.0009	52.0	21268	1.4149	138.1613	304.2098
0.0003	53.0	21677	1.3998	163.519	314.8466
0.0002	54.0	22086	1.4192	141.4035	302.2313
0.0001	55.0	22495	1.4183	150.2878	300.9336
0.0002	56.0	22904	1.4281	172.598	321.0298
0.0018	57.0	23313	1.4229	151.9597	309.6211
0.0009	58.0	23722	1.4263	128.9554	290.2265
0.0003	59.0	24131	1.4430	135.6599	301.7223
0.0002	60.0	24540	1.4487	156.1034	307.6167
0.0004	61.0	24949	1.4252	107.7161	272.8312
0.0001	62.0	25358	1.4254	123.5122	289.272
0.0	63.0	25767	1.4510	121.2901	280.6162
0.0002	64.0	26176	1.4407	111.6482	284.5364
0.0003	65.0	26585	1.4512	123.5207	285.948
0.0006	66.0	26994	1.4476	108.9224	280.1608
0.0005	67.0	27403	1.4721	153.8178	309.4788
0.0004	68.0	27812	1.4675	132.1341	289.9678
0.0001	69.0	28221	1.4712	135.9096	292.8338
0.0001	70.0	28630	1.4712	137.0228	294.8725
0.0	71.0	29039	1.4727	137.9582	292.8438
0.0	72.0	29448	1.4766	135.6514	291.9329
0.0	73.0	29857	1.4808	135.7784	292.3431
0.0	74.0	30266	1.4850	135.5414	291.5527
0.0	75.0	30675	1.4901	134.3224	290.6803
0.0	76.0	31084	1.4943	135.9562	291.6507
0.0	77.0	31493	1.4986	136.0069	291.0294
0.0	78.0	31902	1.5039	139.228	292.1162
0.0	79.0	32311	1.5092	138.6862	291.7796
0.0	80.0	32720	1.5146	139.8375	292.3959
0.0	81.0	33129	1.5208	138.9782	292.097
0.0	82.0	33538	1.5270	140.976	293.3127
0.0	83.0	33947	1.5334	141.3993	292.2359
0.0	84.0	34356	1.5401	141.2258	292.2309
0.0	85.0	34765	1.5472	140.7686	291.4648
0.0	86.0	35174	1.5550	140.6163	291.7997
0.0	87.0	35583	1.5617	142.9104	293.0816
0.0	88.0	35992	1.5700	140.9972	292.0618
0.0	89.0	36401	1.5781	141.5559	292.3054
0.0	90.0	36810	1.5855	142.4109	293.033
0.0	91.0	37219	1.5925	145.0436	293.8586
0.0	92.0	37628	1.6010	144.2648	293.2315
0.0	93.0	38037	1.6083	144.3833	293.3211
0.0	94.0	38446	1.6153	146.3007	294.4095
0.0	95.0	38855	1.6207	146.9864	295.1798
0.0	96.0	39264	1.6269	145.179	293.7054
0.0	97.0	39673	1.6321	148.0107	295.6043
0.0	98.0	40082	1.6358	147.088	295.2686
0.0	99.0	40491	1.6389	148.1503	295.822
0.0	100.0	40900	1.6400	147.3885	295.3741

Framework versions

• Transformers 4.24.0
• Pytorch 1.13.0
• Datasets 2.6.1
• Tokenizers 0.11.0