IoanaLivia/real_data_1_h_synth_wavenet_A_horoscope-whisper-small

This model is a fine-tuned version of openai/whisper-small on the IoanaLivia/real_data_1_h_synth_wavenet_A_horoscope dataset. It achieves the following results on the evaluation set:

• Loss: 0.2701
• Wer: 18.6350

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
0.3223	1.0	46	0.2836	19.9416
0.132	2.0	92	0.2701	18.6350
0.0647	3.0	138	0.2860	18.9789
0.0323	4.0	184	0.3034	19.1336
0.0158	5.0	230	0.3200	19.1680
0.0093	6.0	276	0.3409	18.9789
0.0061	7.0	322	0.3488	18.8757
0.0043	8.0	368	0.3561	19.2367
0.0036	9.0	414	0.3605	18.9960
0.0031	10.0	460	0.3660	19.1680
0.0027	11.0	506	0.3688	19.3571
0.0025	12.0	552	0.3713	19.3055
0.0023	13.0	598	0.3731	19.2711
0.0022	14.0	644	0.3743	19.2883
0.0022	15.0	690	0.3747	19.3227

Framework versions

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