cantillation/Teamim-large-v2_Random-True_DropOut-0.3_Augmented_date-27-06-2024_20-55-08

he-cantillation

This model is a fine-tuned version of ivrit-ai/whisper-v2-pd1-e1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1065
• Wer: 11.5595
• Avg Precision Exact: 0.9158
• Avg Recall Exact: 0.9153
• Avg F1 Exact: 0.9152
• Avg Precision Letter Shift: 0.9349
• Avg Recall Letter Shift: 0.9345
• Avg F1 Letter Shift: 0.9344
• Avg Precision Word Level: 0.9374
• Avg Recall Word Level: 0.9371
• Avg F1 Word Level: 0.9369
• Avg Precision Word Shift: 0.9744
• Avg Recall Word Shift: 0.9751
• Avg F1 Word Shift: 0.9743
• Precision Median Exact: 1.0
• Recall Median Exact: 1.0
• F1 Median Exact: 1.0
• 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: 0.2222
• Recall Min Word Shift: 0.1667
• F1 Min Word Shift: 0.1905

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: 8
• eval_batch_size: 32
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 10000
• 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	8e-05	1	4.8435	119.6859	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
0.1248	0.08	1000	0.1621	23.2816	0.8165	0.8201	0.8175	0.8444	0.8482	0.8454	0.8487	0.8532	0.8501	0.9200	0.9275	0.9227	0.9	0.9	0.8966	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1111	0.0909	0.1111
0.0907	0.16	2000	0.1268	18.2188	0.8624	0.8589	0.8600	0.8863	0.8828	0.8839	0.8905	0.8869	0.8880	0.9507	0.9495	0.9494	0.9231	0.9167	0.9231	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1429	0.125	0.1333
0.04	0.24	3000	0.1182	14.9446	0.8826	0.8813	0.8815	0.9046	0.9035	0.9035	0.9079	0.9067	0.9068	0.9633	0.9635	0.9628	0.9375	0.9375	0.9524	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.125	0.1	0.1176
0.0379	0.32	4000	0.1123	15.0111	0.8839	0.8853	0.8839	0.9081	0.9096	0.9082	0.9110	0.9134	0.9115	0.9615	0.9658	0.9629	0.9333	0.9333	0.9474	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1	0.1	0.1176
0.0274	0.4	5000	0.1092	13.3481	0.9011	0.8989	0.8996	0.9215	0.9194	0.9200	0.9240	0.9225	0.9228	0.9684	0.9685	0.9679	1.0	1.0	0.9630	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1429	0.1	0.1176
0.0242	0.48	6000	0.1078	13.0044	0.9003	0.8977	0.8985	0.9211	0.9187	0.9194	0.9237	0.9216	0.9221	0.9686	0.9681	0.9678	1.0	1.0	0.9630	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1429	0.1	0.1176
0.0134	0.56	7000	0.1039	12.0547	0.9086	0.9072	0.9075	0.9270	0.9257	0.9259	0.9299	0.9288	0.9290	0.9699	0.9701	0.9696	1.0	1.0	1.0	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1538	0.1429	0.1538
0.0118	0.64	8000	0.1077	11.8847	0.9104	0.9110	0.9103	0.9306	0.9313	0.9306	0.9331	0.9339	0.9331	0.9726	0.9748	0.9733	1.0	1.0	1.0	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0769	0.0909	0.0833
0.0084	0.72	9000	0.1071	11.6223	0.9145	0.9141	0.9140	0.9336	0.9332	0.9330	0.9358	0.9355	0.9353	0.9742	0.9751	0.9742	1.0	1.0	1.0	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1429	0.1	0.1176
0.0065	0.8	10000	0.1065	11.5595	0.9158	0.9153	0.9152	0.9349	0.9345	0.9344	0.9374	0.9371	0.9369	0.9744	0.9751	0.9743	1.0	1.0	1.0	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2222	0.1667	0.1905

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.1
• Datasets 2.20.0
• Tokenizers 0.19.1