asr-africa/whisper_DigitalUmuganda_Afrivoice_Shona_10hr_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: 0.9885
• Wer: 43.7776
• Cer: 10.5354

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: 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
1.4704	1.0	221	0.8345	57.1641	12.8241
0.588	2.0	442	0.5671	43.1463	9.0338
0.3483	3.0	663	0.4983	39.4835	8.0809
0.2005	4.0	884	0.4918	38.3288	8.0162
0.1084	5.0	1105	0.5136	38.5771	7.9115
0.0554	6.0	1326	0.5454	38.4405	8.3641
0.0285	7.0	1547	0.5648	38.9744	8.2256
0.0182	8.0	1768	0.5721	38.2915	7.7560
0.012	9.0	1989	0.6069	36.6153	7.7960
0.008	10.0	2210	0.6072	37.8445	7.8022
0.0059	11.0	2431	0.6271	36.4167	7.2819
0.0038	12.0	2652	0.6236	37.1244	7.9361
0.0039	13.0	2873	0.6414	37.1244	7.3650
0.0034	14.0	3094	0.6356	36.3546	7.2603
0.003	15.0	3315	0.6491	36.5284	7.4512
0.0049	16.0	3536	0.6539	38.2915	9.1293
0.0058	17.0	3757	0.6732	38.7758	7.7114
0.0049	18.0	3978	0.6704	37.4224	7.5236
0.0046	19.0	4199	0.6662	37.3231	7.7945
0.0028	20.0	4420	0.6718	36.3422	7.5328
0.002	21.0	4641	0.6796	36.4912	7.5821
0.0019	22.0	4862	0.6730	36.0318	7.5528
0.0021	23.0	5083	0.6833	37.0002	7.6914
0.0018	24.0	5304	0.6928	36.3919	7.4358
0.0018	25.0	5525	0.7052	35.4358	7.4728
0.0014	26.0	5746	0.7111	36.5533	8.0793
0.0019	27.0	5967	0.6961	35.9945	7.2634
0.002	28.0	6188	0.7029	35.6345	7.2942
0.0016	29.0	6409	0.7191	35.7090	7.2649
0.002	30.0	6630	0.7013	35.6593	7.2249
0.0016	31.0	6851	0.7105	35.8704	7.5975
0.0023	32.0	7072	0.7251	35.4482	7.3065
0.0017	33.0	7293	0.7093	35.4482	7.3203
0.0017	34.0	7514	0.7264	36.2180	7.5112
0.0018	35.0	7735	0.7227	35.4731	7.4450

Framework versions

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