shreyasdesaisuperU/whisper-large-large-attempt1

Whisper Large SSD superU

This model is a fine-tuned version of openai/whisper-large on the None dataset. It achieves the following results on the evaluation set:

• Loss: 4.4312
• Wer: 120.4531

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: 16
• eval_batch_size: 8
• 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: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.1192	3.125	100	2.9964	142.5654
1.7195	6.25	200	2.7478	122.2719
0.6545	9.375	300	3.0978	127.3772
0.1503	12.5	400	3.6068	139.0874
0.0827	15.625	500	3.6768	115.5712
0.0556	18.75	600	3.7650	114.9968
0.0441	21.875	700	3.7594	125.3350
0.0346	25.0	800	3.8227	147.7026
0.0205	28.125	900	3.9344	120.8998
0.0166	31.25	1000	3.9918	109.6682
0.0127	34.375	1100	3.9241	109.3491
0.009	37.5	1200	4.1503	110.7211
0.0029	40.625	1300	4.1240	134.5246
0.0007	43.75	1400	4.3018	105.5520
0.0007	46.875	1500	4.3464	106.8283
0.0004	50.0	1600	4.3809	115.5712
0.0003	53.125	1700	4.4061	120.8998
0.0002	56.25	1800	4.4205	120.2936
0.0002	59.375	1900	4.4289	120.4212
0.0002	62.5	2000	4.4312	120.4531

Framework versions

• Transformers 4.46.2
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.20.3