whisper_NCHLT_speech_corpus_Fleurs_Zulu_63hr_v1

This model is a fine-tuned version of openai/whisper-small on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.6630
• Wer: 27.7634
• Cer: 7.5162

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
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.3854	1.0	5509	0.6857	45.5381	10.6552
0.1351	2.0	11018	0.6803	42.7384	10.0055
0.0636	3.0	16527	0.7011	43.5039	10.8099
0.0296	4.0	22036	0.7839	42.6947	9.9912
0.0158	5.0	27545	0.8178	42.1041	10.7433
0.0104	6.0	33054	0.8427	42.4322	10.1411
0.008	7.0	38563	0.8916	41.8198	10.0888
0.0068	8.0	44072	0.9352	45.6037	13.2827
0.0057	9.0	49581	0.9505	41.7760	10.5029
0.0051	10.0	55090	0.9731	43.6133	12.2331
0.0046	11.0	60599	1.0202	41.4261	10.4029
0.0044	12.0	66108	1.0311	43.9414	11.7571
0.0043	13.0	71617	1.0461	41.5573	10.9622
0.0037	14.0	77126	1.0607	41.8416	10.6314
0.0035	15.0	82635	1.0079	41.2948	11.1383
0.003	16.0	88144	1.0468	42.4541	11.7214
0.0031	17.0	93653	1.0365	42.1697	11.0693
0.0027	18.0	99162	1.0952	42.8259	11.5406
0.0026	19.0	104671	1.0987	42.0385	10.5529
0.0024	20.0	110180	1.0835	41.8854	10.9479
0.0025	21.0	115689	1.1063	42.1697	10.7885
0.0023	22.0	121198	1.0948	41.2948	10.3458
0.002	23.0	126707	1.1444	42.7603	11.3406
0.0022	24.0	132216	1.1265	40.8136	10.1054
0.002	25.0	137725	1.1291	41.6667	10.5933
0.002	26.0	143234	1.1695	42.1916	10.5338
0.0021	27.0	148743	1.1100	40.8136	10.6195
0.0016	28.0	154252	1.1380	42.7822	11.4025
0.0014	29.0	159761	1.1595	40.8793	10.3220
0.0015	30.0	165270	1.2079	43.3727	11.5501
0.0015	31.0	170779	1.1418	40.7918	10.8290
0.0013	32.0	176288	1.2209	42.2135	11.0622
0.0014	33.0	181797	1.2364	44.0507	11.5929
0.0014	34.0	187306	1.1969	41.8416	11.1645
0.0012	35.0	192815	1.1686	42.4759	11.4596
0.0013	36.0	198324	1.2171	42.0385	11.1312
0.001	37.0	203833	1.1656	42.6947	11.4287
0.001	38.0	209342	1.1376	41.4042	10.7171
0.0011	39.0	214851	1.1598	41.4917	10.6838
0.001	40.0	220360	1.1863	42.3885	11.6762
0.0008	41.0	225869	1.1719	41.9291	11.8619

Framework versions

• Transformers 4.47.0
• Pytorch 2.1.0+cu118
• Datasets 3.1.0
• Tokenizers 0.21.0