shreyasdesaisuperU/whisper-fine-tune

Whisper Small 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: 3.2716
• Wer: 96.2441

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0	50.0	50	3.7169	94.3662
0.0	100.0	100	3.7773	102.3474
0.0	150.0	150	3.8105	102.3474
0.0	200.0	200	3.9007	101.8779
0.0028	250.0	250	3.2200	90.6103
0.0	300.0	300	3.1754	93.8967
0.0	350.0	350	3.1945	96.2441
0.0	400.0	400	3.2104	96.2441
0.0	450.0	450	3.2225	96.2441
0.0	500.0	500	3.2327	96.2441
0.0	550.0	550	3.2437	96.2441
0.0	600.0	600	3.2501	96.2441
0.0	650.0	650	3.2550	96.2441
0.0	700.0	700	3.2601	96.2441
0.0	750.0	750	3.2634	96.2441
0.0	800.0	800	3.2663	96.2441
0.0	850.0	850	3.2691	96.2441
0.0	900.0	900	3.2717	96.2441
0.0	950.0	950	3.2719	96.2441
0.0	1000.0	1000	3.2716	96.2441

Framework versions

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