whisper-small-Kurdish-Sorani

This model is a fine-tuned version of openai/whisper-small on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1337
• Wer Ortho: 26.7340
• Wer: 24.1171

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: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.3313	0.0248	250	0.4122	63.9017	60.4730
0.2611	0.0496	500	0.3482	55.6466	51.6921
0.2433	0.0744	750	0.3153	53.3507	49.3734
0.2208	0.0992	1000	0.2853	48.3435	44.5947
0.2104	0.1240	1250	0.2717	46.4543	42.7993
0.1828	0.1488	1500	0.2566	45.2822	41.3963
0.1712	0.1736	1750	0.2464	42.9045	39.3076
0.1695	0.1984	2000	0.2390	42.0008	38.2738
0.1709	0.2232	2250	0.2299	40.5875	36.9594
0.1601	0.2480	2500	0.2233	40.0424	36.5079
0.1678	0.2728	2750	0.2194	39.9124	36.4045
0.1534	0.2976	3000	0.2097	37.5954	34.0289
0.1542	0.3224	3250	0.2044	37.0712	33.6069
0.1493	0.3472	3500	0.1993	36.7106	33.3980
0.1258	0.3720	3750	0.1965	36.4086	32.9191
0.1212	0.3968	4000	0.1898	34.8151	31.4823
0.1382	0.4216	4250	0.1867	34.8297	31.4908
0.1368	0.4464	4500	0.1839	34.3244	31.1342
0.1258	0.4712	4750	0.1795	33.8673	30.6553
0.12	0.4960	5000	0.1748	32.6176	29.4126
0.1122	0.5208	5250	0.1699	32.2507	29.0476
0.1191	0.5456	5500	0.1697	32.4394	29.0603
0.1247	0.5704	5750	0.1629	31.2904	28.1826
0.1111	0.5952	6000	0.1633	31.8020	28.6341
0.1163	0.6200	6250	0.1587	30.4600	27.2606
0.0909	0.6448	6500	0.1561	29.8373	26.5433
0.0999	0.6696	6750	0.1534	29.6486	26.6994
0.1224	0.6944	7000	0.1514	29.0762	26.2353
0.0986	0.7192	7250	0.1496	28.8770	26.0707
0.0855	0.7440	7500	0.1479	29.1433	26.3492
0.0866	0.7688	7750	0.1456	28.0424	25.3703
0.0993	0.7937	8000	0.1439	28.1242	25.4040
0.1052	0.8185	8250	0.1414	27.8202	25.1572
0.0853	0.8433	8500	0.1398	27.6482	24.9757
0.0797	0.8681	8750	0.1383	27.0905	24.4188
0.0848	0.8929	9000	0.1375	26.9773	24.2141
0.1011	0.9177	9250	0.1356	26.6229	24.0622
0.0939	0.9425	9500	0.1348	26.7173	24.1803
0.0781	0.9673	9750	0.1341	26.6103	24.0727
0.0894	0.9921	10000	0.1337	26.7340	24.1171

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu118
• Datasets 2.21.0
• Tokenizers 0.19.1