DanaRL/whisper-large-v3-zwksa1305

large-v3-turbo-zwksa1305

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

• Loss: 0.8594
• Wer: 45.6569

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: 8
• eval_batch_size: 4
• 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: 100
• training_steps: 1000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6356	0.8772	100	0.6373	51.2563
0.4797	1.7544	200	0.5798	47.8737
0.3037	2.6316	300	0.5865	45.8821
0.2044	3.5088	400	0.6082	45.5862
0.1288	4.3860	500	0.6568	45.0254
0.0734	5.2632	600	0.7035	45.7717
0.0558	6.1404	700	0.7563	46.1161
0.0367	7.0175	800	0.7771	45.0960
0.0196	7.8947	900	0.8203	46.0720
0.0137	8.7719	1000	0.8594	45.6569

Framework versions

• Transformers 4.44.2
• Pytorch 2.3.1+cu121
• Datasets 3.2.0
• Tokenizers 0.19.1