clt013/whisper-small-ft-malay-peft-epoch-20

Whisper Small FT Malay - CLT013

This model is a fine-tuned version of openai/whisper-small on the Malay Speech 3k dataset. It achieves the following results on the evaluation set:

• Loss: 0.8613

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: 0.001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.1842	0.3731	100	0.8172
0.7488	0.7463	200	0.8014
0.6424	1.1194	300	0.8136
0.5234	1.4925	400	0.7511
0.4951	1.8657	500	0.8203
0.3835	2.2388	600	0.8191
0.3519	2.6119	700	0.8001
0.3868	2.9851	800	0.8011
0.2568	3.3582	900	0.8630
0.2781	3.7313	1000	0.8269
0.2535	4.1045	1100	0.8612
0.2105	4.4776	1200	0.8486
0.2104	4.8507	1300	0.8367
0.1726	5.2239	1400	0.8692
0.1672	5.5970	1500	0.8483
0.1641	5.9701	1600	0.8443
0.1186	6.3433	1700	0.9531
0.1261	6.7164	1800	0.8578
0.1211	7.0896	1900	0.8922
0.0962	7.4627	2000	0.9107
0.1188	7.8358	2100	0.8498
0.0847	8.2090	2200	0.8554
0.0802	8.5821	2300	0.9024
0.0805	8.9552	2400	0.8649
0.0559	9.3284	2500	0.8634
0.053	9.7015	2600	0.8988
0.0555	10.0746	2700	0.8657
0.0415	10.4478	2800	0.8449
0.0401	10.8209	2900	0.8658
0.0318	11.1940	3000	0.8674
0.0245	11.5672	3100	0.8491
0.032	11.9403	3200	0.8694
0.0186	12.3134	3300	0.8620
0.0179	12.6866	3400	0.8555
0.015	13.0597	3500	0.8730
0.0176	13.4328	3600	0.8458
0.0155	13.8060	3700	0.8454
0.0121	14.1791	3800	0.8533
0.0139	14.5522	3900	0.8604
0.009	14.9254	4000	0.8676
0.0095	15.2985	4100	0.8649
0.0059	15.6716	4200	0.8728
0.0065	16.0448	4300	0.8570
0.0049	16.4179	4400	0.8521
0.0042	16.7910	4500	0.8600
0.0051	17.1642	4600	0.8741
0.0037	17.5373	4700	0.8666
0.0037	17.9104	4800	0.8691
0.0029	18.2836	4900	0.8619
0.0023	18.6567	5000	0.8603
0.0019	19.0299	5100	0.8629
0.0018	19.4030	5200	0.8608
0.0018	19.7761	5300	0.8613

Framework versions

• PEFT 0.13.2
• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1