clt013/whisper-large-v3-ft-malay-peft-v1

Whisper Large v3 FT Malay - CLT013

This model is a fine-tuned version of openai/whisper-large-v3 on the Malay Speech 3k dataset. It achieves the following results on the evaluation set:

• Loss: 0.7194

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 3
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.5614	0.0933	25	2.6198
2.9109	0.1866	50	2.5967
2.5414	0.2799	75	2.5518
2.4919	0.3731	100	2.4742
2.5861	0.4664	125	2.3639
2.454	0.5597	150	2.2213
2.32	0.6530	175	2.0616
2.1081	0.7463	200	1.8668
1.7976	0.8396	225	1.6736
1.7597	0.9328	250	1.5280
1.469	1.0261	275	1.4172
1.4484	1.1194	300	1.3275
1.2641	1.2127	325	1.2592
1.1853	1.3060	350	1.1972
1.184	1.3993	375	1.1449
1.1733	1.4925	400	1.0964
1.0707	1.5858	425	1.0568
0.9975	1.6791	450	1.0172
0.9897	1.7724	475	0.9855
1.0223	1.8657	500	0.9524
0.875	1.9590	525	0.9232
0.9242	2.0522	550	0.8968
0.8829	2.1455	575	0.8709
0.8491	2.2388	600	0.8454
0.7793	2.3321	625	0.8236
0.7733	2.4254	650	0.7993
0.7085	2.5187	675	0.7787
0.7403	2.6119	700	0.7596
0.7019	2.7052	725	0.7415
0.722	2.7985	750	0.7309
0.6403	2.8918	775	0.7220
0.699	2.9851	800	0.7194

Framework versions

• PEFT 0.13.0
• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1