VladS159/whisper_small_ro_VladS_02_14_24_4000_steps

Whisper Small Ro - Sarbu Vlad

This model is a fine-tuned version of openai/whisper-small on the Common Voice 16.1 dataset. It achieves the following results on the evaluation set:

• Loss: 0.2920
• Wer: 18.6647

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: 32
• eval_batch_size: 16
• seed: 42
• distributed_type: multi-GPU
• num_devices: 3
• total_train_batch_size: 96
• total_eval_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1437	3.91	500	0.2167	20.5100
0.0268	7.81	1000	0.2202	18.6557
0.008	11.72	1500	0.2478	18.6829
0.0037	15.62	2000	0.2644	18.6708
0.0024	19.53	2500	0.2761	18.6405
0.0018	23.44	3000	0.2844	18.6859
0.0016	27.34	3500	0.2900	18.6799
0.0014	31.25	4000	0.2920	18.6647

Framework versions

• Transformers 4.37.2
• Pytorch 2.2.0
• Datasets 2.17.0
• Tokenizers 0.15.1