finetuning-wav2vec-large-swahili-asr-model_v12

This model is a fine-tuned version of Joshua-Abok/finetuning-wav2vec-large-swahili-asr-model_v10 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3494
• Wer: 0.1436
• Cer: 0.0459

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 0.1
• num_epochs: 30.0

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
13.4755	0.4963	400	7.4714	1.0	1.0
4.5225	0.9926	800	4.3542	1.0	1.0
3.9917	1.4888	1200	3.9714	1.0	1.0
2.8814	1.9851	1600	2.5120	0.8435	0.4681
1.6612	2.4814	2000	1.4443	0.6827	0.2563
1.2274	2.9777	2400	1.0211	0.5581	0.1922
0.9393	3.4739	2800	0.7926	0.4395	0.1422
0.8187	3.9702	3200	0.6590	0.3646	0.1173
0.6702	4.4665	3600	0.5767	0.3271	0.1057
0.6150	4.9628	4000	0.5153	0.2918	0.0917
0.5267	5.4591	4400	0.4795	0.2693	0.0845
0.4953	5.9553	4800	0.4491	0.2495	0.0800
0.4303	6.4516	5200	0.4261	0.2331	0.0727
0.4208	6.9479	5600	0.4069	0.2308	0.0729
0.3791	7.4442	6000	0.3920	0.2139	0.0672
0.3450	7.9404	6400	0.3782	0.2055	0.0648
0.3149	8.4367	6800	0.3690	0.1956	0.0604
0.3126	8.9330	7200	0.3546	0.1895	0.0594
0.2742	9.4293	7600	0.3556	0.1797	0.0568
0.2882	9.9256	8000	0.3474	0.1785	0.0561
0.2461	10.4218	8400	0.3380	0.1729	0.0538
0.2431	10.9181	8800	0.3371	0.1704	0.0547
0.2265	11.4144	9200	0.3389	0.1692	0.0530
0.2264	11.9107	9600	0.3339	0.1675	0.0525
0.1990	12.4069	10000	0.3337	0.1636	0.0527
0.1974	12.9032	10400	0.3335	0.1631	0.0525
0.1762	13.3995	10800	0.3296	0.1594	0.0503
0.1831	13.8958	11200	0.3320	0.1600	0.0498
0.1786	14.3921	11600	0.3338	0.1585	0.0502
0.1702	14.8883	12000	0.3314	0.1567	0.0486
0.1525	15.3846	12400	0.3367	0.1579	0.0500
0.1530	15.8809	12800	0.3372	0.1562	0.0491
0.1502	16.3772	13200	0.3340	0.1512	0.0496
0.1428	16.8734	13600	0.3381	0.1557	0.0486
0.1414	17.3697	14000	0.3443	0.1526	0.0494
0.1434	17.8660	14400	0.3291	0.1508	0.0492
0.1312	18.3623	14800	0.3337	0.1499	0.0474
0.1329	18.8586	15200	0.3380	0.1506	0.0474
0.1278	19.3548	15600	0.3445	0.1477	0.0471
0.1287	19.8511	16000	0.3391	0.1475	0.0467
0.1273	20.3474	16400	0.3412	0.1475	0.0473
0.1211	20.8437	16800	0.3471	0.1479	0.0464
0.1151	21.3400	17200	0.3421	0.1475	0.0463
0.1195	21.8362	17600	0.3442	0.1450	0.0458
0.1160	22.3325	18000	0.3441	0.1466	0.0464
0.1094	22.8288	18400	0.3465	0.1473	0.0468
0.1137	23.3251	18800	0.3499	0.1462	0.0470
0.1122	23.8213	19200	0.3460	0.1462	0.0467
0.1115	24.3176	19600	0.3493	0.1439	0.0460
0.1088	24.8139	20000	0.3520	0.1443	0.0461
0.1077	25.3102	20400	0.3528	0.1458	0.0459
0.1114	25.8065	20800	0.3492	0.1445	0.0459
0.1130	26.3027	21200	0.3447	0.1433	0.0459
0.1034	26.7990	21600	0.3496	0.1448	0.0461
0.1096	27.2953	22000	0.3488	0.1434	0.0457
0.1028	27.7916	22400	0.3498	0.1443	0.0460
0.1072	28.2878	22800	0.3486	0.1431	0.0457
0.1084	28.7841	23200	0.3494	0.1435	0.0456
0.1074	29.2804	23600	0.3497	0.1436	0.0458
0.1117	29.7767	24000	0.3493	0.1439	0.0459
0.1048	30.0	24180	0.3494	0.1436	0.0459

Framework versions

• Transformers 5.6.2
• Pytorch 2.11.0+cu130
• Datasets 4.8.4
• Tokenizers 0.22.2