IoanaLivia/real_data_2_h_synth_standard_A_B_horoscope-whisper-small

This model is a fine-tuned version of openai/whisper-small on the IoanaLivia/real_data_2_h_synth_standard_A_B_horoscope dataset. It achieves the following results on the evaluation set:

• Loss: 0.2835
• Wer: 18.5319

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: 8
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0	0	0.4170	30.1014
0.2648	1.0	66	0.2766	18.9273
0.1007	2.0	132	0.2653	18.8413
0.047	3.0	198	0.2835	18.5319
0.0223	4.0	264	0.3050	18.7382
0.0111	5.0	330	0.3206	19.7696
0.0066	6.0	396	0.3394	19.4946
0.004	7.0	462	0.3476	18.5491
0.003	8.0	528	0.3550	18.5319
0.0027	9.0	594	0.3562	18.6007
0.0021	10.0	660	0.3619	18.6178
0.0019	11.0	726	0.3653	18.7038
0.0018	12.0	792	0.3678	18.7554
0.0016	13.0	858	0.3701	18.8069
0.0016	14.0	924	0.3715	18.8413
0.0015	15.0	990	0.3718	18.7554

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1