whisper-medium-ATCOSIM

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: 0.0993
• Wer: 4.5786

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: 4
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• training_steps: 12500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0448	1.0460	1000	0.0953	6.5544
0.0192	2.0921	2000	0.0926	4.5786
0.0098	3.1381	3000	0.1084	5.3088
0.0079	4.1841	4000	0.0894	4.9609
0.0051	5.2301	5000	0.0971	4.5658
0.0027	6.2762	6000	0.0946	4.2050
0.0023	7.3222	7000	0.1061	4.8406
0.001	8.3682	8000	0.1013	4.6302
0.0006	9.4142	9000	0.1041	4.8535
0.0004	10.4603	10000	0.1057	4.8063
0.0001	11.5063	11000	0.1033	4.6731
0.0002	12.5523	12000	0.0993	4.5786

Framework versions

• Transformers 4.43.3
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1