whisper-swi-asr_new2

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

• Loss: 0.1832
• Wer: 0.0890

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: 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: 50
• training_steps: 16000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3675	0.2219	1000	0.3479	0.2085
0.2826	0.4439	2000	0.2770	0.1661
0.2485	0.6658	3000	0.2378	0.1863
0.2316	0.8877	4000	0.2140	0.1715
0.1103	1.1096	5000	0.2161	0.1470
0.1064	1.3316	6000	0.2054	0.1268
0.1011	1.5535	7000	0.2017	0.1141
0.0934	1.7754	8000	0.1868	0.1268
0.0968	1.9973	9000	0.1811	0.1054
0.0389	2.2193	10000	0.1925	0.1208
0.0384	2.4412	11000	0.1878	0.1029
0.0346	2.6631	12000	0.1789	0.1081
0.0304	2.8850	13000	0.1755	0.1078
0.0131	3.1070	14000	0.1814	0.0890
0.0098	3.3289	15000	0.1837	0.1062
0.0079	3.5508	16000	0.1832	0.0890

Framework versions

• Transformers 4.43.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1