whisper-large-v3-pt-cv19-fleurs-ct2

This model is a fine-tuned version of openai/whisper-large-v3 on the fsicoli/cv19-fleurs default dataset converted to CT2. It achieves the following results on the evaluation set:

• Loss: 0.1823
• Wer: 0.0756

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: 6.25e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• 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: 10000
• training_steps: 50000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0559	2.2883	5000	0.1096	0.0730
0.0581	4.5767	10000	0.1326	0.0829
0.0225	6.8650	15000	0.1570	0.0849
0.0088	9.1533	20000	0.1704	0.0840
0.0065	11.4416	25000	0.1823	0.0849
0.006	13.7300	30000	0.1808	0.0809
0.0055	16.0183	35000	0.1811	0.0790
0.0031	18.3066	40000	0.1907	0.0784
0.0011	20.5950	45000	0.1852	0.0771
0.0003	22.8833	50000	0.1848	0.0756

Framework versions

• Transformers 4.45.0.dev0
• Pytorch 2.4.1
• Datasets 2.21.0
• Tokenizers 0.19.1