asr-africa/whisper_DigitalUmuganda_Afrivoice_Shona_1hr_v1

Whisper Small Shona - Beijuka Bruno

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

• Loss: 1.5125
• Wer: 69.0996
• Cer: 17.2441

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: 16
• 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: 500
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.448	1.0	12	2.3789	133.5673	53.7368
2.3262	2.0	24	2.3298	124.4924	46.1703
2.2746	3.0	36	2.2040	116.1695	39.4590
2.0498	4.0	48	2.0031	105.1642	35.9980
1.8194	5.0	60	1.8268	105.3648	35.4961
1.6512	6.0	72	1.6481	90.0727	25.6863
1.4776	7.0	84	1.4833	84.3068	21.7946
1.1776	8.0	96	1.3361	78.4407	20.5831
1.0071	9.0	108	1.2316	73.6275	17.8038
0.9545	10.0	120	1.1512	69.1903	16.2732
0.7517	11.0	132	1.0924	68.8142	16.8805
0.6386	12.0	144	1.0463	65.6556	15.1205
0.5353	13.0	156	1.0123	63.9759	14.8045
0.4123	14.0	168	0.9882	62.3214	14.5318
0.3863	15.0	180	0.9721	60.9927	14.4079
0.2775	16.0	192	0.9743	59.7393	13.0786
0.1919	17.0	204	0.9750	58.6362	12.8741
0.1377	18.0	216	0.9922	59.9148	13.0229
0.0891	19.0	228	0.9891	60.4412	14.5969
0.0658	20.0	240	1.0039	59.5137	13.6395
0.0381	21.0	252	1.0140	58.7365	13.6271
0.0264	22.0	264	1.0197	57.1823	12.1646
0.0172	23.0	276	1.0256	56.8062	13.3017
0.0116	24.0	288	1.0402	57.4079	12.3660
0.0091	25.0	300	1.0421	57.4831	13.6488
0.0074	26.0	312	1.0482	56.7059	12.5922
0.0059	27.0	324	1.0552	55.7032	11.8207
0.0048	28.0	336	1.0683	56.3550	12.5333
0.004	29.0	348	1.0692	56.1544	12.4682
0.0037	30.0	360	1.0769	58.0847	14.0237
0.0032	31.0	372	1.0818	57.7087	13.1809
0.0028	32.0	384	1.0870	58.0847	14.7177
0.0025	33.0	396	1.0921	57.8842	13.8285
0.0024	34.0	408	1.0983	57.3828	13.1344
0.0021	35.0	420	1.0996	58.4106	14.8758
0.0019	36.0	432	1.1043	55.4525	11.8207
0.0017	37.0	444	1.1086	57.7338	13.9834
0.0017	38.0	456	1.1142	58.3605	14.0082
0.0016	39.0	468	1.1178	55.7032	12.3908
0.0014	40.0	480	1.1238	55.9789	13.2212
0.0013	41.0	492	1.1261	57.8591	14.5349
0.0012	42.0	504	1.1276	57.2825	13.0167
0.0012	43.0	516	1.1304	55.1767	11.7122
0.0011	44.0	528	1.1342	54.9260	11.6255
0.001	45.0	540	1.1380	54.8759	11.7401
0.001	46.0	552	1.1406	54.7255	11.7184
0.0009	47.0	564	1.1436	54.4497	11.6286
0.0009	48.0	576	1.1475	55.0013	12.3815
0.0008	49.0	588	1.1494	54.7756	11.6657
0.0008	50.0	600	1.1525	54.3495	11.6069
0.0007	51.0	612	1.1562	54.1740	11.6905
0.0007	52.0	624	1.1561	56.5555	13.0291
0.0007	53.0	636	1.1573	56.4302	12.9609
0.0007	54.0	648	1.1602	54.5751	11.6843
0.0007	55.0	660	1.1618	55.2519	12.8401
0.0006	56.0	672	1.1629	54.4497	11.6286
0.0006	57.0	684	1.1655	54.3495	11.6936
0.0006	58.0	696	1.1673	54.8759	12.5643
0.0006	59.0	708	1.1684	54.1990	11.6781
0.0006	60.0	720	1.1699	54.5751	12.4249
0.0006	61.0	732	1.1719	56.7059	13.0415

Framework versions

• Transformers 4.44.2
• Pytorch 2.1.0+cu118
• Datasets 3.0.0
• Tokenizers 0.19.1