whisper-large-v3-turbo-persian-audiodata2

This model is a fine-tuned version of openai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.1979
• Wer: 42.4096

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: 0.001
• train_batch_size: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 50
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
2.0420	1.0	5	1.2280	44.8996
1.9990	2.0	10	1.1840	45.0602
1.8969	3.0	15	1.1200	43.2932
1.7966	4.0	20	1.0419	43.7751
1.5880	5.0	25	0.9716	42.4096
1.4181	6.0	30	0.9191	38.8755
1.2844	7.0	35	0.8705	36.0643
1.0876	8.0	40	0.8371	38.0723
0.9473	9.0	45	0.8437	41.0442
0.9232	10.0	50	0.8159	38.7149
0.9232	26.0	52	0.8091	38.7952
0.9232	27.0	54	0.8043	39.3574
0.5928	28.0	56	0.8029	40.1606
0.5928	29.0	58	0.7987	38.3133
0.4714	30.0	60	0.8038	37.0281
0.4714	31.0	62	0.8098	36.3052
0.4714	32.0	64	0.8083	36.7871
0.4555	33.0	66	0.8136	38.7149
0.4555	34.0	68	0.8217	39.4378
0.4328	35.0	70	0.8310	38.6345
0.4328	36.0	72	0.8371	37.5904
0.4328	37.0	74	0.8482	37.9116
0.4061	38.0	76	0.8510	37.9116
0.4061	39.0	78	0.8545	40.0803
0.2561	40.0	80	0.8577	40.4819
0.2561	41.0	82	0.8642	40.8032
0.2561	42.0	84	0.8782	39.5181
0.3224	43.0	86	0.8928	40.9639
0.3224	44.0	88	0.9002	41.7671
0.2982	45.0	90	0.9020	40.6426
0.2982	46.0	92	0.9061	42.1687
0.2982	47.0	94	0.9262	41.4458
0.2625	48.0	96	0.9401	41.6867
0.2625	49.0	98	0.9421	40.1606
0.2262	50.0	100	0.9564	43.8554
0.2262	51.0	102	0.9636	45.7028
0.3451	21.0	105	0.9795	40.8835
0.2854	22.0	110	1.0011	46.1847
0.2876	23.0	115	0.9650	42.2490
0.2733	24.0	120	0.9405	39.9197
0.2690	25.0	125	1.0017	41.6867
0.2461	26.0	130	0.9712	39.0361
0.2338	27.0	135	0.9927	39.9197
0.2520	28.0	140	1.0054	41.8474
0.2184	29.0	145	0.9881	40.0
0.2064	30.0	150	0.9915	40.5622
0.1959	31.0	155	1.0218	40.6426
0.1867	32.0	160	1.0504	40.4819
0.1774	33.0	165	1.0113	39.7590
0.1821	34.0	170	1.0357	39.9197
0.1561	35.0	175	1.0310	40.4819
0.1429	36.0	180	1.0668	42.4096
0.1307	37.0	185	1.0383	41.8474
0.1147	38.0	190	1.0794	41.0442
0.1106	39.0	195	1.0610	40.8835
0.1093	40.0	200	1.0934	41.6867
0.0982	41.0	205	1.0682	40.6426
0.0879	42.0	210	1.0832	39.7590
0.0855	43.0	215	1.1129	40.8835
0.0716	44.0	220	1.1060	42.4096
0.0799	45.0	225	1.0980	41.0442
0.0729	46.0	230	1.1165	41.2048
0.0690	47.0	235	1.1068	40.4016
0.0622	48.0	240	1.1066	40.2410
0.0604	49.0	245	1.0999	42.0884
0.0532	50.0	250	1.1072	42.0884
0.0507	51.0	255	1.1105	41.7671
0.0449	52.0	260	1.1121	42.2490
0.0402	53.0	265	1.1475	41.6867
0.0374	54.0	270	1.1638	41.4458
0.0366	55.0	275	1.1283	42.4096
0.0349	56.0	280	1.1476	41.8474
0.0331	57.0	285	1.1761	41.2851
0.0293	58.0	290	1.1688	41.5261
0.0317	59.0	295	1.1367	41.2048
0.0272	60.0	300	1.1569	41.2048
0.0268	61.0	305	1.1896	42.2490
0.0251	62.0	310	1.1667	42.4900
0.0243	63.0	315	1.1719	42.1687
0.0252	64.0	320	1.1886	44.1767
0.0244	65.0	325	1.1842	42.5703
0.0223	66.0	330	1.1770	42.0080
0.0210	67.0	335	1.1844	42.4096
0.0205	68.0	340	1.1825	42.2490
0.0201	69.0	345	1.1772	42.5703
0.0200	70.0	350	1.1825	41.7671
0.0204	71.0	355	1.1822	41.6867
0.0195	72.0	360	1.1885	42.9719
0.0195	73.0	365	1.1874	42.8112
0.0187	74.0	370	1.1843	42.3293
0.0183	75.0	375	1.1847	43.4538
0.0182	76.0	380	1.1908	42.5703
0.0170	77.0	385	1.1942	42.3293
0.0170	78.0	390	1.1903	42.3293
0.0176	79.0	395	1.1954	42.6506
0.0174	80.0	400	1.1910	42.8112
0.0179	81.0	405	1.1938	42.7309
0.0173	82.0	410	1.1978	43.0522
0.0162	83.0	415	1.1926	44.0964
0.0165	84.0	420	1.1940	43.3735
0.0172	85.0	425	1.1941	43.3735
0.0167	86.0	430	1.1949	42.8916
0.0162	87.0	435	1.1968	43.9357
0.0164	88.0	440	1.1966	43.0522
0.0164	89.0	445	1.1975	42.8112
0.0169	90.0	450	1.1993	42.5703
0.0164	91.0	455	1.1969	42.7309
0.0154	92.0	460	1.1978	42.3293
0.0159	93.0	465	1.1974	42.3293
0.0163	94.0	470	1.1999	41.7671
0.0161	95.0	475	1.1958	42.0884
0.0153	96.0	480	1.1976	42.0884
0.0162	97.0	485	1.1961	42.6506
0.0167	98.0	490	1.1988	42.1687
0.0159	99.0	495	1.1985	42.6506
0.0164	100.0	500	1.1979	42.4096

Framework versions

• PEFT 0.19.1
• Transformers 5.13.1
• Pytorch 2.11.0+cu128
• Datasets 4.0.0
• Tokenizers 0.22.2