fs-w-xavier-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: 0.3811
• Wer: 91.5954
• Cer: 70.8949

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
3.9903	4.5872	500	4.0338	98.0532	78.4954
1.0456	9.1743	1000	1.2616	95.8215	77.5850
0.4233	13.7615	1500	0.6630	103.7037	79.1309
0.3371	18.3486	2000	0.5218	99.8575	75.2576
0.3005	22.9358	2500	0.4897	106.1728	83.6225
0.2556	27.5229	3000	0.4328	94.6819	73.4799
0.2132	32.1101	3500	0.4141	97.1510	74.5964
0.1848	36.6972	4000	0.3955	93.1624	71.2384
0.1635	41.2844	4500	0.3883	94.6819	72.8358
0.139	45.8716	5000	0.3811	91.5954	70.8949

Framework versions

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