psst-model-8e-1s-hp400hz

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

• Loss: 0.5231
• Wer: 0.1535
• Iu F1: 0.7274
• Iu Tp: 834
• Iu Fp: 461
• Iu Fn: 164

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 664
• training_steps: 9480

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Iu F1	Iu Tp	Iu Fp	Iu Fn
0.9693	1.0	1185	0.4934	0.2967	0.5833	464	551	112
0.6271	2.0	2370	0.4776	0.2406	0.7606	429	123	147
0.3445	3.0	3555	0.4993	0.2225	0.7678	458	159	118
0.1631	4.0	4740	0.5507	0.2293	0.7564	455	172	121
0.0622	5.0	5925	0.6060	0.2221	0.7671	466	173	110
0.0276	6.0	7110	0.6858	0.2240	0.7574	448	159	128
0.0098	7.0	8295	0.7383	0.2164	0.7672	445	139	131
0.0032	8.0	9480	0.7756	0.2149	0.7679	440	130	136

Framework versions

• Transformers 5.6.2
• Pytorch 2.6.0+cu124
• Datasets 2.21.0
• Tokenizers 0.22.2