BA_Model_V3

This model is a fine-tuned version of openai/whisper-large-v2 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3495
• Wer: 21.1224
• Cer: 12.2080

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-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: cosine_with_restarts
• lr_scheduler_warmup_ratio: 0.2
• num_epochs: 8
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
1.0358	1.0	278	0.6770	23.3686	13.9275
0.2753	2.0	556	0.3313	20.7071	12.0535
0.2109	3.0	834	0.3098	20.7204	12.1752
0.1603	4.0	1112	0.3129	20.4645	11.8444
0.1224	5.0	1390	0.3242	20.8034	12.0535
0.0956	6.0	1668	0.3353	20.7802	11.9894
0.0781	7.0	1946	0.3464	21.0659	12.1725
0.0716	8.0	2224	0.3495	21.1224	12.2080

Framework versions

• Transformers 4.38.2
• Pytorch 2.2.1+cu121
• Datasets 2.18.0
• Tokenizers 0.15.2