Makkoen/whisper-large-v3-turbo-cit-do01-wd0-lr3e-06-FULL4

./4607

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the 4607 FULL-2024-09-26 dataset. It achieves the following results on the evaluation set:

• Loss: 0.5166
• Wer Ortho: 28.3540
• Wer: 20.6411

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
0.8514	0.7715	200	0.6094	33.7638	25.0150
0.6006	1.5429	400	0.5527	30.1201	22.1503
0.5069	2.3144	600	0.5322	28.9055	21.1352
0.4291	3.0858	800	0.5234	27.9422	20.1002
0.3772	3.8573	1000	0.5151	28.0120	20.3539
0.3449	4.6287	1200	0.5166	28.3540	20.6411

Framework versions

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