whisper-large-v2-ca-5awr

This model is a fine-tuned version of openai/whisper-large-v2 on the cymen-arfor/15awr train[:33%] main dataset. It achieves the following results on the evaluation set:

• Loss: 1.1840
• Wer: 0.4096

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: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 5000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0354	9.4340	1000	0.8619	0.4279
0.0073	18.8679	2000	1.0045	0.4271
0.0008	28.3019	3000	1.0738	0.4097
0.0003	37.7358	4000	1.1648	0.4106
0.0002	47.1698	5000	1.1840	0.4096

Framework versions

• Transformers 4.44.0
• Pytorch 2.4.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1