shreyasdesaisuperU/whisper-medium-attempt2-1000-orders

Whisper Medium with 1000 orders 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: 3.0551
• Wer: 114.5969

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
3.3178	0.3831	100	3.1949	211.3345
2.5484	0.7663	200	2.3725	319.0767
2.0593	1.1494	300	2.2722	248.3126
2.0603	1.5326	400	2.2479	253.4519
2.0258	1.9157	500	2.2259	302.7082
1.6141	2.2989	600	2.2450	191.7982
1.729	2.6820	700	2.2441	221.9182
1.4741	3.0651	800	2.3114	169.6906
1.351	3.4483	900	2.3248	199.7299
1.4636	3.8314	1000	2.3174	177.5240
1.1092	4.2146	1100	2.4755	143.9276
1.0713	4.5977	1200	2.4954	122.6198
1.0927	4.9808	1300	2.4714	153.5291
0.7693	5.3640	1400	2.6916	124.3843
0.7594	5.7471	1500	2.7017	125.2438
0.576	6.1303	1600	2.8599	117.2104
0.5304	6.5134	1700	2.9010	120.1431
0.508	6.8966	1800	2.9175	120.5536
0.3794	7.2797	1900	3.0556	111.9378
0.3816	7.6628	2000	3.0551	114.5969

Framework versions

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