clt013/whisper-large-v3-ft-malay-peft-epoch-20

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: 0.4149

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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
2.6257	0.3731	100	2.4258
2.3394	0.7463	200	1.7629
1.5256	1.1194	300	1.1983
1.0648	1.4925	400	0.9111
0.7458	1.8657	500	0.5754
0.5287	2.2388	600	0.5316
1.0221	2.6119	700	1.6308
0.9045	2.9851	800	0.4655
0.3766	3.3582	900	0.4186
0.3767	3.7313	1000	0.4082
0.355	4.1045	1100	0.4022
0.3284	4.4776	1200	0.3996
0.3229	4.8507	1300	0.3928
0.318	5.2239	1400	0.3925
0.3083	5.5970	1500	0.3905
0.3035	5.9701	1600	0.3900
0.2792	6.3433	1700	0.3897
0.2948	6.7164	1800	0.3872
0.2537	7.0896	1900	0.3869
0.2703	7.4627	2000	0.3907
0.2696	7.8358	2100	0.3891
0.2672	8.2090	2200	0.3923
0.2645	8.5821	2300	0.3900
0.2419	8.9552	2400	0.3894
0.2361	9.3284	2500	0.3928
0.2361	9.7015	2600	0.3915
0.2229	10.0746	2700	0.3922
0.2151	10.4478	2800	0.3962
0.248	10.8209	2900	0.3945
0.2152	11.1940	3000	0.3981
0.2144	11.5672	3100	0.4012
0.2278	11.9403	3200	0.3970
0.2017	12.3134	3300	0.4002
0.216	12.6866	3400	0.4018
0.2166	13.0597	3500	0.4013
0.2176	13.4328	3600	0.4043
0.2002	13.8060	3700	0.4054
0.1752	14.1791	3800	0.4066
0.1892	14.5522	3900	0.4048
0.2121	14.9254	4000	0.4078
0.209	15.2985	4100	0.4069
0.1857	15.6716	4200	0.4079
0.1867	16.0448	4300	0.4097
0.1833	16.4179	4400	0.4093
0.1918	16.7910	4500	0.4130
0.1815	17.1642	4600	0.4123
0.1854	17.5373	4700	0.4135
0.181	17.9104	4800	0.4131
0.1756	18.2836	4900	0.4130
0.177	18.6567	5000	0.4143
0.1829	19.0299	5100	0.4137
0.1809	19.4030	5200	0.4163
0.1777	19.7761	5300	0.4149

Framework versions

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