psst-portuguese-4epochs

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: 0.3566
• Wer: 0.1550
• Iu Accuracy: 0.9423
• Iu Precision: 0.6807
• Iu Recall: 0.8749
• Iu F1: 0.7657

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: 393
• training_steps: 5612

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Iu Accuracy	Iu Precision	Iu Recall	Iu F1
1.0364	0.5005	702	0.5356	0.2971	0.9459	0.8371	0.7049	0.7653
0.9917	1.0007	1404	0.5033	0.2596	0.9349	0.5191	0.8472	0.6438
0.6096	1.5012	2106	0.4928	0.2446	0.9408	0.7770	0.8229	0.7993
0.5511	2.0014	2808	0.4806	0.2225	0.9456	0.8397	0.7274	0.7795
0.2754	2.5020	3510	0.5112	0.2295	0.9453	0.8460	0.7535	0.7971
0.2526	3.0021	4212	0.4893	0.2237	0.9429	0.8374	0.7691	0.8018
0.1129	3.5027	4914	0.5588	0.2180	0.9451	0.8284	0.7795	0.8032
0.1118	4.0	5612	0.5561	0.2202	0.9420	0.8217	0.8003	0.8109

Framework versions

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