muk-luganda-digits-classification

This model is a fine-tuned version of facebook/wav2vec2-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 3.1383
• Accuracy: 0.4118

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: 4
• eval_batch_size: 1
• 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: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
2.3979	1.0	68	2.3931	0.1176
2.3557	2.0	136	2.3405	0.1765
2.2886	3.0	204	2.3076	0.2353
2.2136	4.0	272	2.2603	0.1471
2.1519	5.0	340	2.2511	0.2059
2.0177	6.0	408	2.1587	0.2353
1.9045	7.0	476	2.2249	0.3529
1.763	8.0	544	2.2783	0.2647
1.6427	9.0	612	2.0562	0.3824
1.5392	10.0	680	2.0552	0.3529
1.3602	11.0	748	1.9803	0.3529
1.3406	12.0	816	1.9426	0.4412
1.2032	13.0	884	2.0370	0.3529
1.0842	14.0	952	1.9971	0.5
1.0615	15.0	1020	1.8678	0.4412
0.9674	16.0	1088	1.8230	0.4706
0.808	17.0	1156	1.9590	0.4412
0.7434	18.0	1224	2.0104	0.4412
0.6936	19.0	1292	2.1243	0.4118
0.5685	20.0	1360	1.9530	0.4706
0.5631	21.0	1428	2.0039	0.4118
0.4919	22.0	1496	2.3431	0.4118
0.4903	23.0	1564	2.5384	0.4412
0.4504	24.0	1632	2.3131	0.4412
0.361	25.0	1700	2.6200	0.3824
0.3683	26.0	1768	2.6182	0.4118
0.3205	27.0	1836	2.5689	0.4412
0.2389	28.0	1904	2.6241	0.4706
0.1786	29.0	1972	2.7555	0.4412
0.1627	30.0	2040	2.6370	0.5
0.1764	31.0	2108	3.3250	0.3529
0.2364	32.0	2176	2.9069	0.4412
0.1772	33.0	2244	2.9289	0.4118
0.1929	34.0	2312	3.1186	0.4118
0.2075	35.0	2380	3.1905	0.4118
0.1517	36.0	2448	3.0900	0.4118
0.149	37.0	2516	3.1103	0.3824
0.1156	38.0	2584	3.1738	0.4412
0.2291	39.0	2652	3.1478	0.4412
0.1768	40.0	2720	3.1383	0.4118

Framework versions

• Transformers 4.57.0
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1