cantillation/Teamim-IvritAI-large-v3-turbo_WeightDecay-0.005_Augmented_WithSRT_date-15-04-2025

he-cantillation

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

• Loss: 0.4220
• Wer: 35.5610
• Avg Precision Exact: 0.4895
• Avg Recall Exact: 0.5009
• Avg F1 Exact: 0.4943
• Avg Precision Letter Shift: 0.5105
• Avg Recall Letter Shift: 0.5234
• Avg F1 Letter Shift: 0.5157
• Avg Precision Word Level: 0.5228
• Avg Recall Word Level: 0.5364
• Avg F1 Word Level: 0.5279
• Avg Precision Word Shift: 0.6946
• Avg Recall Word Shift: 0.7228
• Avg F1 Word Shift: 0.7055
• Precision Median Exact: 0.4524
• Recall Median Exact: 0.4833
• F1 Median Exact: 0.4706
• Precision Max Exact: 1.0
• Recall Max Exact: 1.0
• F1 Max Exact: 1.0
• Precision Min Exact: 0.0
• Recall Min Exact: 0.0
• F1 Min Exact: 0.0
• Precision Min Letter Shift: 0.0
• Recall Min Letter Shift: 0.0
• F1 Min Letter Shift: 0.0
• Precision Min Word Level: 0.0
• Recall Min Word Level: 0.0
• F1 Min Word Level: 0.0
• Precision Min Word Shift: 0.0
• Recall Min Word Shift: 0.0
• F1 Min Word Shift: 0.0

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: 2
• seed: 42
• optimizer: Use OptimizerNames.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: 1000
• training_steps: 60000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Avg Precision Exact	Avg Recall Exact	Avg F1 Exact	Avg Precision Letter Shift	Avg Recall Letter Shift	Avg F1 Letter Shift	Avg Precision Word Level	Avg Recall Word Level	Avg F1 Word Level	Avg Precision Word Shift	Avg Recall Word Shift	Avg F1 Word Shift	Precision Median Exact	Recall Median Exact	F1 Median Exact	Precision Max Exact	Recall Max Exact	F1 Max Exact	Precision Min Exact	Recall Min Exact	F1 Min Exact	Precision Min Letter Shift	Recall Min Letter Shift	F1 Min Letter Shift	Precision Min Word Level	Recall Min Word Level	F1 Min Word Level	Precision Min Word Shift	Recall Min Word Shift	F1 Min Word Shift
No log	0.0001	1	7.2503	108.6320	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0	0	0	0	0	0	0	0	0	0	0	0	0	0	0
0.1342	0.2962	2500	0.6879	55.3095	0.3378	0.3531	0.3445	0.3736	0.3930	0.3815	0.3902	0.4089	0.3972	0.5839	0.6279	0.6015	0.1935	0.2174	0.2021	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0737	0.5925	5000	0.6062	50.9023	0.3793	0.3990	0.3878	0.4084	0.4316	0.4184	0.4244	0.4481	0.4344	0.6099	0.6620	0.6313	0.25	0.2772	0.2647	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0637	0.8887	7500	0.5195	47.3558	0.3950	0.4133	0.4024	0.4216	0.4437	0.4298	0.4354	0.4590	0.4441	0.6160	0.6641	0.6339	0.2581	0.2917	0.2692	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0499	1.1850	10000	0.5286	46.2515	0.4063	0.4231	0.4125	0.4327	0.4530	0.4401	0.4477	0.4708	0.4561	0.6263	0.6728	0.6433	0.2667	0.3	0.2817	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0468	1.4812	12500	0.5122	46.7592	0.4081	0.4204	0.4130	0.4345	0.4489	0.4401	0.4476	0.4627	0.4534	0.6198	0.6532	0.6324	0.2571	0.2806	0.2667	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.053	1.7775	15000	0.4899	45.7365	0.4067	0.4211	0.4119	0.4348	0.4524	0.4408	0.4490	0.4689	0.4556	0.6308	0.6698	0.6455	0.2727	0.2934	0.2778	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0357	2.0737	17500	0.4765	44.8087	0.4227	0.4343	0.4277	0.4494	0.4636	0.4554	0.4654	0.4795	0.4709	0.6496	0.6805	0.6614	0.2857	0.3095	0.2934	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0452	2.3699	20000	0.4817	43.5380	0.4225	0.4342	0.4263	0.4483	0.4632	0.4533	0.4617	0.4777	0.4666	0.6398	0.6742	0.6514	0.2642	0.2857	0.2689	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0384	2.6662	22500	0.4465	42.9034	0.4410	0.4517	0.4450	0.4656	0.4779	0.4700	0.4783	0.4919	0.4829	0.6630	0.6879	0.6717	0.3134	0.3333	0.32	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0171	2.9624	25000	0.4247	39.2373	0.4574	0.4690	0.4619	0.4819	0.4952	0.4868	0.4948	0.5104	0.5006	0.6782	0.7077	0.6892	0.3433	0.3703	0.3504	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0142	3.2587	27500	0.4705	43.7539	0.4300	0.4472	0.4361	0.4545	0.4748	0.4613	0.4665	0.4887	0.4739	0.6422	0.6844	0.6568	0.3103	0.3333	0.3220	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0163	3.5549	30000	0.4480	41.7772	0.4505	0.4613	0.4548	0.4759	0.4873	0.4798	0.4900	0.5018	0.4937	0.6693	0.6940	0.6777	0.3455	0.375	0.3523	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0081	3.8512	32500	0.4588	40.7224	0.4470	0.4565	0.4506	0.4701	0.4810	0.4741	0.4829	0.4946	0.4872	0.6673	0.6950	0.6775	0.3333	0.35	0.3418	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0156	4.1474	35000	0.4382	38.0863	0.4597	0.4695	0.4639	0.4835	0.4956	0.4882	0.4964	0.5099	0.5012	0.6763	0.7053	0.6873	0.3563	0.3846	0.3704	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0146	4.4437	37500	0.4453	39.8763	0.4662	0.4771	0.4704	0.4896	0.5038	0.4946	0.5019	0.5173	0.5072	0.6714	0.7033	0.6826	0.3690	0.4118	0.3795	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0075	4.7399	40000	0.4473	37.8747	0.4659	0.4760	0.4700	0.4880	0.5008	0.4930	0.5004	0.5147	0.5059	0.6739	0.7035	0.6852	0.3732	0.4064	0.3867	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0054	5.0361	42500	0.4328	37.4954	0.4664	0.4775	0.4711	0.4898	0.5035	0.4952	0.5020	0.5167	0.5074	0.6801	0.7111	0.6920	0.3611	0.3934	0.3743	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0019	5.3324	45000	0.4505	38.5108	0.4614	0.4718	0.4656	0.4856	0.4984	0.4907	0.4985	0.5121	0.5037	0.6791	0.7063	0.6895	0.3606	0.3901	0.3709	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0089	5.6286	47500	0.4472	38.6348	0.4593	0.4684	0.4628	0.4828	0.4941	0.4871	0.4949	0.5074	0.4995	0.6734	0.7023	0.6841	0.3438	0.3624	0.3530	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0171	5.9249	50000	0.4119	36.0337	0.4815	0.4914	0.4854	0.5048	0.5163	0.5092	0.5169	0.5291	0.5214	0.6935	0.7229	0.7049	0.4286	0.4535	0.4407	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0025	6.2211	52500	0.4380	36.5982	0.4804	0.4927	0.4853	0.5030	0.5172	0.5086	0.5136	0.5294	0.5197	0.6862	0.7181	0.6985	0.4068	0.4367	0.4205	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0062	6.5174	55000	0.4319	37.3904	0.4784	0.4895	0.4832	0.5006	0.5141	0.5063	0.5129	0.5269	0.5185	0.6844	0.7136	0.6958	0.4142	0.4390	0.4255	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0085	6.8136	57500	0.4211	36.6566	0.4856	0.4969	0.4903	0.5068	0.5195	0.5118	0.5188	0.5329	0.5239	0.6872	0.7160	0.6977	0.4353	0.4721	0.4552	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0048	7.1098	60000	0.4220	35.5610	0.4895	0.5009	0.4943	0.5105	0.5234	0.5157	0.5228	0.5364	0.5279	0.6946	0.7228	0.7055	0.4524	0.4833	0.4706	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu126
• Datasets 2.12.0
• Tokenizers 0.20.1