psst-model-4e-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: 0.3002
• Wer: 0.1543
• Iu F1: 0.7271
• Iu Tp: 858
• Iu Fp: 504
• Iu Fn: 140

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
1.0854	0.4998	592	0.5175	0.2750	0.7104	390	132	186
0.9392	0.9996	1184	0.4803	0.2940	0.6760	458	321	118
0.5928	1.4989	1776	0.4777	0.2403	0.7141	356	65	220
0.5330	1.9987	2368	0.4661	0.2474	0.7615	463	177	113
0.2883	2.4981	2960	0.4986	0.2358	0.7579	446	155	130
0.2779	2.9979	3552	0.4922	0.2152	0.7717	463	161	113
0.0953	3.4973	4144	0.5633	0.2202	0.7687	447	140	129
0.0886	3.9970	4736	0.5514	0.2187	0.7615	447	151	129
0.0886	4.0	4740	0.5514	0.2187	0.7615	447	151	129

Framework versions

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