psst-model-4e-1s-lp3200hz

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: 0.3469
• Wer: 0.1555
• Iu F1: 0.7211
• Iu Tp: 835
• Iu Fp: 483
• Iu Fn: 163

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: 332
• training_steps: 4740

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Iu F1	Iu Tp	Iu Fp	Iu Fn
1.1125	0.4998	592	0.5420	0.2591	0.7102	402	154	174
0.9877	0.9996	1184	0.4973	0.2566	0.7547	460	183	116
0.6032	1.4989	1776	0.4974	0.2439	0.7410	389	85	187
0.5663	1.9987	2368	0.4861	0.2359	0.7508	470	206	106
0.3061	2.4981	2960	0.5101	0.2450	0.7572	446	156	130
0.2972	2.9979	3552	0.5068	0.2263	0.7663	459	163	117
0.0988	3.4973	4144	0.5757	0.2254	0.7643	428	116	148
0.1003	3.9970	4736	0.5681	0.2247	0.7654	447	145	129
0.1003	4.0	4740	0.5681	0.2248	0.7665	448	145	128

Framework versions

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