whisper-audio-abuse-feature

This model is a fine-tuned version of parambharat/whisper-tiny-south-indic on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4947
• Accuracy: 0.8174
• Macro F1-score: 0.7845

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Macro F1-score
7.8152	0.52	10	7.7606	0.0	0.0
7.6445	1.04	20	7.4368	0.6718	0.1348
7.1349	1.56	30	6.6932	0.6897	0.4082
6.2832	2.08	40	5.7185	0.6897	0.4082
5.3726	2.6	50	4.8531	0.6897	0.4082
4.4541	3.12	60	4.1519	0.6897	0.4082
3.9045	3.64	70	3.5417	0.6897	0.4082
3.3278	4.16	80	2.9984	0.6897	0.4082
2.7361	4.68	90	2.5167	0.6897	0.4082
2.3838	5.19	100	2.1039	0.6897	0.4082
1.9557	5.71	110	1.7366	0.6897	0.4082
1.5922	6.23	120	1.4170	0.6897	0.4082
1.3228	6.75	130	1.1497	0.7068	0.4669
1.0767	7.27	140	0.9322	0.7779	0.6610
0.8649	7.79	150	0.7814	0.7860	0.6800
0.7058	8.31	160	0.6608	0.8165	0.7453
0.6253	8.83	170	0.5688	0.8327	0.7936
0.5269	9.35	180	0.5137	0.8291	0.7896
0.5016	9.87	190	0.4862	0.8300	0.7747
0.4409	10.39	200	0.4776	0.8040	0.7796
0.3793	10.91	210	0.4511	0.8345	0.8048
0.3501	11.43	220	0.4491	0.8228	0.7922
0.3692	11.95	230	0.4254	0.8327	0.7982
0.3148	12.47	240	0.4452	0.8228	0.7901
0.3114	12.99	250	0.4543	0.8345	0.7905
0.2812	13.51	260	0.4398	0.8363	0.7992
0.2635	14.03	270	0.4607	0.8327	0.7960
0.2491	14.55	280	0.4818	0.8327	0.7864
0.2825	15.06	290	0.4616	0.8219	0.7920
0.2494	15.58	300	0.4784	0.8129	0.7897
0.2127	16.1	310	0.4669	0.8273	0.7862
0.2003	16.62	320	0.4760	0.8174	0.7796
0.1907	17.14	330	0.4845	0.8246	0.7943
0.177	17.66	340	0.4870	0.8219	0.7867
0.1877	18.18	350	0.4884	0.8201	0.7909
0.1511	18.7	360	0.4907	0.8228	0.7845
0.1826	19.22	370	0.4932	0.8165	0.7839
0.1419	19.74	380	0.4947	0.8174	0.7845

Framework versions

• Transformers 4.33.0
• Pytorch 2.0.0
• Datasets 2.1.0
• Tokenizers 0.13.3