nomnoos37/stt-turbo-250314-v2.0.5

Whisper Turbo ko

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the custom dataset. It achieves the following results on the evaluation set:

• Loss: 0.2173

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: 0.001
• train_batch_size: 64
• eval_batch_size: 256
• seed: 42
• 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
• lr_scheduler_warmup_steps: 100
• training_steps: 300
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.806	0.0474	10	1.5761
0.555	0.0948	20	1.1182
0.2932	0.1422	30	0.9172
0.2319	0.1896	40	0.7350
0.1991	0.2370	50	0.6569
0.1782	0.2844	60	0.5887
0.1676	0.3318	70	0.5292
0.1526	0.3791	80	0.4856
0.1464	0.4265	90	0.4509
0.1429	0.4739	100	0.4312
0.1417	0.5213	110	0.3998
0.1359	0.5687	120	0.3802
0.1279	0.6161	130	0.3608
0.125	0.6635	140	0.3369
0.111	0.7109	150	0.3189
0.1203	0.7583	160	0.3088
0.1416	0.8057	170	0.3010
0.1135	0.8531	180	0.2795
0.108	0.9005	190	0.2734
0.0881	0.9479	200	0.2702
0.1225	0.9953	210	0.2580
0.0874	1.0427	220	0.2523
0.0997	1.0900	230	0.2610
0.0964	1.1374	240	0.2388
0.082	1.1848	250	0.2300
0.0793	1.2322	260	0.2279
0.0816	1.2796	270	0.2243
0.0885	1.3270	280	0.2198
0.0745	1.3744	290	0.2180
0.077	1.4218	300	0.2173

Framework versions

• PEFT 0.14.0
• Transformers 4.47.1
• Pytorch 2.3.1+cu121
• Datasets 3.0.0
• Tokenizers 0.21.0