CngFSt3T_experimental_model

This model is a fine-tuned version of openai/whisper-medium on the Marcusxx/CngFSt3T dataset. It achieves the following results on the evaluation set:

• Loss: 0.0896
• Cer: 213.5728

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: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• training_steps: 10000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
0.3967	2.2321	1000	0.4536	272.8711
0.0776	4.4643	2000	0.1882	227.7485
0.0132	6.6964	3000	0.1046	263.5602
0.0055	8.9286	4000	0.0888	243.2638
0.0012	11.1607	5000	0.0862	245.4492
0.001	13.3929	6000	0.0832	223.3024
0.0011	15.625	7000	0.0859	207.0502
0.0004	17.8571	8000	0.0879	203.3827
0.0003	20.0893	9000	0.0892	220.0368
0.0003	22.3214	10000	0.0896	213.5728

Framework versions

• Transformers 4.41.2
• Pytorch 2.2.2+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1