Makkoen/whisper-large-v3-cit-do01-wd0-lr3e-06-steps1200-FULL5

./7326

This model is a fine-tuned version of openai/whisper-large-v3 on the 7326 FULL-2024-10-24 dataset. It achieves the following results on the evaluation set:

• Loss: 0.3926
• Wer Ortho: 22.5695
• Wer: 15.5891

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: 3e-06
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• 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: 300
• training_steps: 1200
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.6853	0.4851	200	0.4600	25.9249	18.6707
0.5251	0.9703	400	0.4211	24.1878	17.0180
0.4314	1.4554	600	0.4028	23.3234	16.1387
0.4047	1.9406	800	0.3950	23.0530	16.0798
0.361	2.4257	1000	0.3948	23.0407	15.9424
0.3441	2.9109	1200	0.3926	22.5695	15.5891

Framework versions

• Transformers 4.45.1
• Pytorch 1.13.1+cu117
• Datasets 3.0.1
• Tokenizers 0.20.0