asr-africa/whisper_NCHLT_speech_corpus_Xhosa_5hr_v1

Whisper Small Xhosa - Beijuka Bruno

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

• Loss: 1.6980
• Model Preparation Time: 0.0077
• Wer: 68.2920
• Cer: 30.3879

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	Model Preparation Time	Wer	Cer
0.8602	1.0	671	0.5255	0.0077	41.0282	8.8247
0.1398	2.0	1342	0.4073	0.0077	31.3360	7.2621
0.0391	3.0	2013	0.3998	0.0077	27.0138	6.4473
0.015	4.0	2684	0.3925	0.0077	26.7518	6.1563
0.0081	5.0	3355	0.4090	0.0077	24.4597	5.5205
0.0045	6.0	4026	0.4028	0.0077	23.8376	5.5429
0.0029	7.0	4697	0.4145	0.0077	23.6739	5.5518
0.0024	8.0	5368	0.4073	0.0077	23.2482	5.4936
0.0024	9.0	6039	0.4387	0.0077	24.0013	5.5876
0.0034	10.0	6710	0.4386	0.0077	24.7872	5.9861
0.004	11.0	7381	0.4363	0.0077	23.4447	5.4265
0.0027	12.0	8052	0.4621	0.0077	23.0190	5.5966
0.0021	13.0	8723	0.4396	0.0077	23.5102	5.6638
0.0048	14.0	9394	0.4492	0.0077	24.0341	5.8608
0.0033	15.0	10065	0.4337	0.0077	22.3314	5.4578
0.002	16.0	10736	0.4442	0.0077	23.1827	5.5787
0.0024	17.0	11407	0.4648	0.0077	22.8225	5.6056
0.0037	18.0	12078	0.4663	0.0077	24.3615	6.1428
0.0022	19.0	12749	0.4539	0.0077	24.0013	7.9158
0.0018	20.0	13420	0.4558	0.0077	21.7420	5.4802
0.0008	21.0	14091	0.4777	0.0077	22.5606	5.4488
0.0032	22.0	14762	0.4593	0.0077	22.7898	5.9906
0.0018	23.0	15433	0.4592	0.0077	24.2305	8.0949
0.002	24.0	16104	0.4815	0.0077	23.7394	6.2145
0.0019	25.0	16775	0.5052	0.0077	21.6110	5.6503
0.0008	26.0	17446	0.4684	0.0077	22.7570	6.9129
0.0003	27.0	18117	0.4699	0.0077	21.2181	5.3772
0.0024	28.0	18788	0.4757	0.0077	23.7721	5.9145
0.0017	29.0	19459	0.4717	0.0077	22.0367	5.3459
0.0009	30.0	20130	0.4779	0.0077	23.4119	5.6951
0.0009	31.0	20801	0.4776	0.0077	22.6588	5.7085
0.0022	32.0	21472	0.4808	0.0077	23.6084	5.8428
0.0015	33.0	22143	0.5002	0.0077	23.7394	6.0309
0.0013	34.0	22814	0.4833	0.0077	22.4623	5.7041
0.001	35.0	23485	0.4900	0.0077	22.4951	5.4265
0.0006	36.0	24156	0.4907	0.0077	22.3641	5.4847
0.0004	37.0	24827	0.4819	0.0077	21.9057	5.5071

Framework versions

• Transformers 4.47.1
• Pytorch 2.1.0+cu118
• Datasets 3.2.0
• Tokenizers 0.21.0