cantillation/Teamim-small_Random_WeightDecay-0.05_Augmented_Old-Data_date-21-07-2024_14-33

he-cantillation

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

• Loss: 0.4208
• Wer: 18.5421
• Avg Precision Exact: 0.8482
• Avg Recall Exact: 0.8623
• Avg F1 Exact: 0.8545
• Avg Precision Letter Shift: 0.8778
• Avg Recall Letter Shift: 0.8926
• Avg F1 Letter Shift: 0.8844
• Avg Precision Word Level: 0.8803
• Avg Recall Word Level: 0.8946
• Avg F1 Word Level: 0.8866
• Avg Precision Word Shift: 0.9416
• Avg Recall Word Shift: 0.9549
• Avg F1 Word Shift: 0.9475
• Precision Median Exact: 0.9167
• Recall Median Exact: 0.9167
• F1 Median Exact: 0.9167
• 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: 0.6154
• Recall Min Word Shift: 0.6667
• F1 Min Word Shift: 0.64

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: 500
• training_steps: 200000
• 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	7.8969	120.8943	0.0002	0.0000	0.0000	0.0030	0.0028	0.0027	0.0024	0.0137	0.0027	0.0141	0.0233	0.0133	0.0	0.0	0.0	0.0625	0.0074	0.0132	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0
0.0777	0.8	10000	0.2194	22.7167	0.7902	0.7985	0.7935	0.8205	0.8298	0.8243	0.8264	0.8358	0.8303	0.9175	0.9318	0.9237	0.9091	0.9091	0.8889	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
0.0363	1.6	20000	0.2639	19.9619	0.8266	0.8330	0.8289	0.8554	0.8623	0.8580	0.8620	0.8676	0.8640	0.9419	0.9519	0.9459	0.9091	0.9167	0.9091	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6154	0.6429	0.64
0.0189	2.4	30000	0.2936	20.0678	0.8200	0.8284	0.8233	0.8485	0.8574	0.8521	0.8538	0.8621	0.8571	0.9365	0.9491	0.9419	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.4167	0.4545	0.4348
0.0089	3.2	40000	0.3115	19.8347	0.8154	0.8264	0.8200	0.8456	0.8570	0.8504	0.8488	0.8601	0.8536	0.9396	0.9526	0.9452	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.5	0.5556	0.5263
0.0095	4.0	50000	0.3330	20.0042	0.8325	0.8425	0.8366	0.8628	0.8739	0.8674	0.8668	0.8780	0.8715	0.9376	0.9481	0.9418	0.9091	0.9167	0.9091	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.5385	0.5833	0.5600
0.0062	4.8	60000	0.3605	19.6228	0.8291	0.8399	0.8336	0.8619	0.8733	0.8667	0.8651	0.8752	0.8693	0.9376	0.9466	0.9413	0.9091	0.9167	0.9167	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	0.0833	0.0952
0.0052	5.6	70000	0.3855	20.4492	0.8137	0.8244	0.8182	0.8398	0.8513	0.8448	0.8446	0.8568	0.8499	0.9330	0.9448	0.9380	0.9091	0.9091	0.9	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.4545	0.5556	0.5
0.0018	6.4	80000	0.3755	20.1950	0.8275	0.8378	0.8318	0.8579	0.8695	0.8628	0.8617	0.8732	0.8666	0.9402	0.9511	0.9447	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6667	0.6	0.6316
0.0038	7.2	90000	0.3829	20.4281	0.8308	0.8429	0.8359	0.8586	0.8715	0.8640	0.8619	0.8738	0.8669	0.9328	0.9467	0.9387	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0833	0.0769	0.08
0.0029	8.0	100000	0.3823	19.4533	0.8423	0.8504	0.8455	0.8686	0.8773	0.8720	0.8716	0.8795	0.8747	0.9435	0.9498	0.9457	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6154	0.6667	0.64
0.0013	8.8	110000	0.3992	19.7499	0.8323	0.8449	0.8378	0.8609	0.8745	0.8668	0.8639	0.8767	0.8694	0.9404	0.9532	0.9459	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6154	0.625	0.625
0.0015	9.6	120000	0.3963	19.8135	0.8317	0.8426	0.8364	0.8629	0.8745	0.8679	0.8660	0.8775	0.8710	0.9355	0.9460	0.9399	0.9167	0.9167	0.9167	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
0.0005	10.4	130000	0.4010	19.3261	0.8398	0.8500	0.8441	0.8705	0.8816	0.8751	0.8737	0.8845	0.8782	0.9379	0.9497	0.9430	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2727	0.2727	0.2727
0.0003	11.2	140000	0.4015	19.3685	0.8386	0.8518	0.8444	0.8676	0.8817	0.8738	0.8714	0.8844	0.8771	0.9319	0.9461	0.9381	0.9091	0.9167	0.9091	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.2727	0.2727	0.2727
0.0007	12.0	150000	0.4071	19.2837	0.8341	0.8455	0.8389	0.8642	0.8765	0.8695	0.8684	0.8802	0.8735	0.9380	0.9493	0.9428	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.5833	0.5385	0.5600
0.0008	12.8	160000	0.4111	19.3685	0.8391	0.8491	0.8433	0.8683	0.8791	0.8729	0.8722	0.8824	0.8765	0.9373	0.9494	0.9425	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.5833	0.5833	0.5833
0.0002	13.6	170000	0.4107	19.4745	0.8295	0.8400	0.8339	0.8607	0.8721	0.8655	0.8645	0.8751	0.8689	0.9379	0.9493	0.9427	0.9091	0.9167	0.9091	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.5833	0.5833	0.5833
0.0001	14.4	180000	0.4214	18.7752	0.8473	0.8609	0.8533	0.8780	0.8925	0.8844	0.8810	0.8940	0.8867	0.9407	0.9539	0.9464	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6364	0.7273	0.6957
0.0001	15.2	190000	0.4146	18.7116	0.8474	0.8609	0.8533	0.8768	0.8910	0.8831	0.8804	0.8944	0.8866	0.9413	0.9550	0.9473	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6667	0.7273	0.6957
0.0005	16.0	200000	0.4208	18.5421	0.8482	0.8623	0.8545	0.8778	0.8926	0.8844	0.8803	0.8946	0.8866	0.9416	0.9549	0.9475	0.9167	0.9167	0.9167	1.0	1.0	1.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.6154	0.6667	0.64

Framework versions

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