psst-model-2e-1s-augmented

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: 0.4813
• Wer: 0.1560
• Iu F1: 0.7139
• Iu Tp: 842
• Iu Fp: 519
• Iu Fn: 156

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: 663
• training_steps: 9476

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Iu F1	Iu Tp	Iu Fp	Iu Fn
0.7848	0.2499	1184	0.5144	0.2582	0.7123	437	214	139
0.5081	0.4998	2368	0.5032	0.2775	0.7409	449	187	127
0.3347	0.7498	3552	0.5189	0.2290	0.7591	446	153	130
0.1791	0.9997	4736	0.5520	0.2268	0.7658	448	146	128
0.0744	1.2495	5920	0.5993	0.2288	0.7607	445	149	131
0.0425	1.4994	7104	0.6511	0.2224	0.7556	405	91	171
0.0158	1.7493	8288	0.7028	0.2159	0.7629	420	105	156
0.0085	1.9993	9472	0.7248	0.2166	0.7774	447	127	129
0.0085	2.0	9476	0.7247	0.2163	0.7781	447	126	129

Framework versions

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