shreyasdesaisuperU/whisper-small-names-3000

Whisper Small 3000 Names SSD superU

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

• Loss: 0.1550
• Wer: 38.3523

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
0.6702	0.2392	100	0.5894	64.7727
0.4536	0.4785	200	0.4488	55.3977
0.4115	0.7177	300	0.4097	50.2841
0.3396	0.9569	400	0.3204	45.7386
0.2729	1.1962	500	0.3147	44.3182
0.2448	1.4354	600	0.3061	45.4545
0.2398	1.6746	700	0.2533	44.6023
0.2017	1.9139	800	0.2488	45.7386
0.1189	2.1531	900	0.2282	43.1818
0.1214	2.3923	1000	0.2319	43.1818
0.1156	2.6316	1100	0.2012	41.4773
0.1031	2.8708	1200	0.1863	42.3295
0.0704	3.1100	1300	0.1746	38.9205
0.0796	3.3493	1400	0.1622	40.0568
0.0868	3.5885	1500	0.1624	39.7727
0.075	3.8278	1600	0.1620	39.7727
0.0423	4.0670	1700	0.1543	37.7841
0.0458	4.3062	1800	0.1507	38.6364
0.0378	4.5455	1900	0.1530	38.0682
0.0379	4.7847	2000	0.1550	38.3523

Framework versions

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