cantillation/Teamim-medium_Random_WeightDecay-0.005_Augmented_New-Data_date-06-03-2025

he-cantillation

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

• Loss: 0.1114
• Wer: 17.1345
• Avg Precision Exact: 0.6531
• Avg Recall Exact: 0.6641
• Avg F1 Exact: 0.6580
• Avg Precision Letter Shift: 0.6616
• Avg Recall Letter Shift: 0.6724
• Avg F1 Letter Shift: 0.6665
• Avg Precision Word Level: 0.6676
• Avg Recall Word Level: 0.6769
• Avg F1 Word Level: 0.6717
• Avg Precision Word Shift: 0.8665
• Avg Recall Word Shift: 0.8803
• Avg F1 Word Shift: 0.8727
• Precision Median Exact: 0.8330
• Recall Median Exact: 0.8604
• F1 Median Exact: 0.8545
• 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: 0.2
• Recall Min Word Shift: 0.1429
• F1 Min Word Shift: 0.1667

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: 8
• 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: 500
• training_steps: 2000
• 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
3.4463	0.0620	100	3.2418	104.1520	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
2.0028	0.1240	200	1.7998	96.7251	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
0.9391	0.1860	300	0.9484	88.9474	0.0784	0.0913	0.0839	0.1095	0.1251	0.1163	0.1455	0.1667	0.1546	0.3273	0.4719	0.3849	0.08	0.0931	0.0851	0.2353	0.25	0.2424	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0769	0.1667	0.1176
0.786	0.2480	400	0.7401	81.8129	0.1117	0.1166	0.1137	0.1342	0.1400	0.1365	0.1591	0.1649	0.1612	0.3595	0.4136	0.3816	0.1076	0.1077	0.1076	0.4138	0.4286	0.4211	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.5189	0.3100	500	0.5562	66.4912	0.3001	0.3145	0.3066	0.3201	0.3348	0.3268	0.3340	0.3454	0.3391	0.4716	0.5012	0.4844	0.3478	0.3618	0.3529	0.6667	0.7222	0.6364	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0714	0.0769	0.0741
0.3767	0.3720	600	0.4395	55.7310	0.3506	0.3479	0.3482	0.3735	0.3711	0.3712	0.3813	0.3780	0.3785	0.5910	0.5980	0.5922	0.4097	0.4226	0.4259	0.8	0.8421	0.8205	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1	0.0741	0.0851
0.3089	0.4340	700	0.3508	44.0936	0.4179	0.4286	0.4227	0.4360	0.4470	0.4410	0.4433	0.4523	0.4473	0.6664	0.6895	0.6769	0.5256	0.5455	0.5375	0.8276	0.85	0.8136	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.15	0.125	0.1364
0.2818	0.4960	800	0.2893	37.6023	0.4850	0.5021	0.4928	0.5045	0.5217	0.5123	0.5119	0.5256	0.5181	0.7129	0.7333	0.7221	0.6036	0.6364	0.6191	0.9231	0.9048	0.9057	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.2341	0.5580	900	0.2547	32.6901	0.5273	0.5442	0.5351	0.5457	0.5628	0.5536	0.5531	0.5687	0.5603	0.7361	0.7568	0.7455	0.6190	0.65	0.6266	0.9310	0.9310	0.9310	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.1905	0.1538	0.1702
0.1907	0.6200	1000	0.2243	31.7544	0.5673	0.5839	0.5749	0.5784	0.5948	0.5858	0.5848	0.5969	0.5902	0.7580	0.7782	0.7673	0.6878	0.7083	0.7042	0.9583	0.9583	0.9583	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0556	0.0526	0.0541
0.1923	0.6820	1100	0.2019	28.4795	0.5495	0.5558	0.5524	0.5664	0.5728	0.5693	0.5728	0.5786	0.5754	0.7993	0.8032	0.7997	0.6878	0.6927	0.6948	0.9565	0.9565	0.9565	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2381	0.1724	0.2000
0.1822	0.7440	1200	0.1793	27.1930	0.6045	0.6202	0.6117	0.6204	0.6361	0.6276	0.6255	0.6377	0.6311	0.7823	0.7982	0.7895	0.72	0.755	0.7310	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	0.1034	0.1200
0.1721	0.8060	1300	0.1672	24.1520	0.6115	0.6306	0.6203	0.6203	0.6400	0.6294	0.6250	0.6426	0.6331	0.8062	0.8274	0.8160	0.755	0.8006	0.7755	0.9655	1.0	0.9825	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.15	0.1034	0.1224
0.1573	0.8679	1400	0.1528	21.3450	0.6228	0.6391	0.6303	0.6346	0.6508	0.6421	0.6381	0.6532	0.6450	0.8210	0.8365	0.8280	0.7643	0.8155	0.8074	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.04	0.0370	0.0385
0.1521	0.9299	1500	0.1354	20.0	0.6349	0.6502	0.6420	0.6459	0.6618	0.6533	0.6498	0.6637	0.6562	0.8392	0.8552	0.8464	0.7713	0.8038	0.8082	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.15	0.1034	0.1224
0.1492	0.9919	1600	0.1303	18.7719	0.6603	0.6764	0.6678	0.6703	0.6868	0.6780	0.6750	0.6899	0.6819	0.8516	0.8677	0.8590	0.8165	0.8297	0.8165	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2	0.1481	0.1702
0.1072	1.0539	1700	0.1243	18.1871	0.6240	0.6377	0.6303	0.6320	0.6461	0.6385	0.6388	0.6515	0.6446	0.8351	0.8490	0.8414	0.8	0.8192	0.8156	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0714	0.0741	0.0727
0.0983	1.1159	1800	0.1174	18.0117	0.6380	0.6529	0.6449	0.6463	0.6611	0.6532	0.6511	0.6633	0.6566	0.8593	0.8745	0.8661	0.8209	0.8297	0.8284	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2	0.1379	0.1633
0.0916	1.1779	1900	0.1122	17.8363	0.6601	0.6733	0.6662	0.6698	0.6829	0.6758	0.6746	0.6863	0.6799	0.8585	0.8734	0.8652	0.8261	0.8604	0.8444	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2	0.1429	0.1667
0.0854	1.2399	2000	0.1114	17.1345	0.6531	0.6641	0.6580	0.6616	0.6724	0.6665	0.6676	0.6769	0.6717	0.8665	0.8803	0.8727	0.8330	0.8604	0.8545	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2	0.1429	0.1667

Framework versions

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