shreyasdesaisuperU/whisper-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: 4.2685
• Wer: 166.6349

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• 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
4.1121	3.125	100	3.5671	154.6120
2.6613	6.25	200	2.8860	158.7150
1.8679	9.375	300	2.8342	143.7977
1.1096	12.5	400	3.0283	167.7163
0.563	15.625	500	3.2773	167.3982
0.2032	18.75	600	3.4815	167.4618
0.0899	21.875	700	3.6164	151.9720
0.0431	25.0	800	3.7659	154.4211
0.0262	28.125	900	3.8327	188.4860
0.0264	31.25	1000	3.8547	173.1234
0.0118	34.375	1100	3.9458	184.9237
0.0076	37.5	1200	4.0480	178.3079
0.0036	40.625	1300	4.1518	159.7964
0.0014	43.75	1400	4.1739	164.6310
0.0011	46.875	1500	4.2014	173.6641
0.001	50.0	1600	4.2262	147.2646
0.001	53.125	1700	4.2510	159.1921
0.0009	56.25	1800	4.2570	168.0025
0.0009	59.375	1900	4.2650	166.7621
0.0008	62.5	2000	4.2685	166.6349

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1