whisper-finetuned-v3_2e

This model is a fine-tuned version of openai/whisper-large-v3-turbo on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0648
• Wer: 60.0

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 4
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.6639	0.1120	100	1.1985	235.1515
0.9978	0.2240	200	0.6341	150.1299
0.4115	0.3359	300	0.2679	88.0952
0.26	0.4479	400	0.2050	83.2468
0.2076	0.5599	500	0.1975	77.7922
0.1776	0.6719	600	0.1711	70.8658
0.1589	0.7839	700	0.1450	74.8485
0.1366	0.8959	800	0.1362	68.0519
0.124	1.0078	900	0.1124	62.1645
0.0944	1.1198	1000	0.1125	63.8528
0.0841	1.2318	1100	0.1039	61.6017
0.0759	1.3438	1200	0.0972	75.1082
0.0773	1.4558	1300	0.0892	60.4329
0.0709	1.5677	1400	0.0806	58.3983
0.0632	1.6797	1500	0.0813	58.0519
0.0613	1.7917	1600	0.0745	56.9697
0.0603	1.9037	1700	0.0689	55.1515
0.0477	2.0157	1800	0.0683	54.0693
0.0294	2.1277	1900	0.0660	60.4329
0.029	2.2396	2000	0.0648	60.0

Framework versions

• Transformers 4.51.3
• Pytorch 2.7.0+cu126
• Datasets 3.5.1
• Tokenizers 0.21.1