Kushtrim/whisper-large-v3-turbo-sq-test3

Whisper Large V3 Turbo SQ

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the Common Voice 19.0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3161
• Wer: 23.9627

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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
• num_epochs: 4
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5057	0.5112	500	0.5311	39.0968
0.3303	1.0225	1000	0.4321	34.5439
0.3165	1.5337	1500	0.3782	31.1893
0.1799	2.0450	2000	0.3470	27.7212
0.1945	2.5562	2500	0.3320	26.4628
0.1277	3.0675	3000	0.3235	24.8606
0.1502	3.5787	3500	0.3161	23.9627

Framework versions

• Transformers 4.45.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.1