multiclass-fz-enc-base-en

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

• Loss: 0.1255
• Wer: 16.9991
• Cer: 6.8791

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: 8
• seed: 42
• 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	Cer
0.1012	4.5872	500	0.2311	21.6999	9.7647
0.0068	9.1743	1000	0.1444	24.1216	11.7228
0.0008	13.7615	1500	0.1342	17.8063	7.2913
0.0005	18.3486	2000	0.1309	17.9012	7.3600
0.0003	22.9358	2500	0.1286	17.5689	7.2484
0.0002	27.5229	3000	0.1275	17.6638	7.2655
0.0002	32.1101	3500	0.1266	16.9516	6.8619
0.0002	36.6972	4000	0.1260	16.9516	6.8705
0.0001	41.2844	4500	0.1255	16.9991	6.8533
0.0001	45.8716	5000	0.1255	16.9991	6.8791

Framework versions

• Transformers 4.45.1
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0