psst-model-8e-1s-hp600hz

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: 0.5230
• Wer: 0.1545
• Iu F1: 0.7132
• Iu Tp: 802
• Iu Fp: 449
• Iu Fn: 196

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
0.9730	1.0	1185	0.4946	0.2584	0.7377	464	218	112
0.6296	2.0	2370	0.4878	0.2588	0.7502	413	112	163
0.3408	3.0	3555	0.5017	0.2308	0.7480	457	189	119
0.1610	4.0	4740	0.5694	0.2291	0.7625	451	156	125
0.0614	5.0	5925	0.6215	0.2266	0.7622	444	145	132
0.0283	6.0	7110	0.6950	0.2212	0.7707	447	137	129
0.0096	7.0	8295	0.7433	0.2207	0.7730	446	132	130
0.0029	8.0	9480	0.7838	0.2145	0.7730	441	124	135

Framework versions

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