COPAS-whisper-lg-3-Nov29

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

• Loss: 0.1596
• Wer: 13.1363

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: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9253	2.3529	100	0.3717	26.4368
0.1676	4.7059	200	0.2410	21.4559
0.04	7.0588	300	0.2064	17.6245
0.0204	9.4118	400	0.1890	18.1171
0.0092	11.7647	500	0.1929	17.8982
0.0051	14.1176	600	0.1794	21.3465
0.0062	16.4706	700	0.1682	14.8331
0.0046	18.8235	800	0.1798	15.2709
0.0026	21.1765	900	0.1717	15.3257
0.0022	23.5294	1000	0.1805	14.7236
0.0019	25.8824	1100	0.1961	15.4351
0.0027	28.2353	1200	0.1614	22.1128
0.0024	30.5882	1300	0.1653	17.4603
0.0023	32.9412	1400	0.1850	17.6793
0.0023	35.2941	1500	0.1564	13.1910
0.0014	37.6471	1600	0.1551	13.8478
0.002	40.0	1700	0.1533	13.5194
0.0013	42.3529	1800	0.1496	13.4100
0.0007	44.7059	1900	0.1555	13.4100
0.0006	47.0588	2000	0.1524	13.3005
0.0001	49.4118	2100	0.1526	13.4647
0.0001	51.7647	2200	0.1540	13.3005
0.0001	54.1176	2300	0.1549	13.2458
0.0001	56.4706	2400	0.1555	13.2458
0.0001	58.8235	2500	0.1561	13.3552
0.0001	61.1765	2600	0.1565	13.3005
0.0	63.5294	2700	0.1569	13.2458
0.0	65.8824	2800	0.1573	13.1910
0.0	68.2353	2900	0.1576	13.1910
0.0	70.5882	3000	0.1579	13.1910
0.0	72.9412	3100	0.1581	13.3552
0.0	75.2941	3200	0.1584	13.4100
0.0	77.6471	3300	0.1586	13.4100
0.0	80.0	3400	0.1588	12.7531
0.0	82.3529	3500	0.1589	13.0816
0.0	84.7059	3600	0.1591	12.6984
0.0	87.0588	3700	0.1593	13.1363
0.0	89.4118	3800	0.1594	13.1363
0.0	91.7647	3900	0.1594	13.0816
0.0	94.1176	4000	0.1595	13.0816
0.0	96.4706	4100	0.1596	13.0816
0.0	98.8235	4200	0.1596	13.1363

Framework versions

• Transformers 4.43.4
• Pytorch 2.4.1
• Datasets 3.0.0
• Tokenizers 0.19.1