shreyasdesaisuperU/whisper-medium-attempt1-with-silence-text

Whisper Medium with Silence text SSD superU

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

• Loss: 3.9457
• Wer: 108.2931

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
2.502	3.125	100	2.8539	127.1068
1.3556	6.25	200	2.4154	115.9689
0.7175	9.375	300	2.7462	102.9791
0.2532	12.5	400	3.1023	114.7880
0.1002	15.625	500	3.2100	117.9817
0.0519	18.75	600	3.2355	128.6098
0.0341	21.875	700	3.3208	105.7166
0.024	25.0	800	3.5071	102.5497
0.015	28.125	900	3.6051	103.0327
0.0098	31.25	1000	3.6116	118.4917
0.004	34.375	1100	3.6429	102.4960
0.0053	37.5	1200	3.6545	105.0188
0.0018	40.625	1300	3.7910	113.3924
0.0005	43.75	1400	3.8308	110.8964
0.0004	46.875	1500	3.8879	105.2872
0.0005	50.0	1600	3.9086	103.7574
0.0003	53.125	1700	3.9241	110.1181
0.0003	56.25	1800	3.9369	107.2732
0.0003	59.375	1900	3.9431	108.2662
0.0003	62.5	2000	3.9457	108.2931

Framework versions

• Transformers 4.46.2
• Pytorch 2.2.2+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3