ap-t0fmdpXB2NbEJd7JOHgXSX

This model is a fine-tuned version of openai/whisper-large-v3-turbo on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4376
• Model Preparation Time: 0.0075
• Wer: 0.1362

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: 3e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.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: 400
• num_epochs: 24
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer
0.4535	1.0	126	0.2399	0.0075	0.1887
0.155	2.0	252	0.2657	0.0075	0.2152
0.1266	3.0	378	0.3336	0.0075	0.1854
0.1164	4.0	504	0.3666	0.0075	0.2016
0.0896	5.0	630	0.4160	0.0075	0.2116
0.0676	6.0	756	0.4054	0.0075	0.1876
0.053	7.0	882	0.3995	0.0075	0.1854
0.0433	8.0	1008	0.4179	0.0075	0.1847
0.0378	9.0	1134	0.3901	0.0075	0.2229
0.0318	10.0	1260	0.4534	0.0075	0.1928
0.0228	11.0	1386	0.4582	0.0075	0.2006
0.0184	12.0	1512	0.4295	0.0075	0.1766
0.015	13.0	1638	0.4408	0.0075	0.1753
0.0096	14.0	1764	0.4205	0.0075	0.1584
0.0082	15.0	1890	0.4330	0.0075	0.1545
0.007	16.0	2016	0.4251	0.0075	0.1543
0.0037	17.0	2142	0.4135	0.0075	0.1585
0.0021	18.0	2268	0.4260	0.0075	0.1406
0.001	19.0	2394	0.4281	0.0075	0.1382
0.0006	20.0	2520	0.4253	0.0075	0.1375
0.0003	21.0	2646	0.4347	0.0075	0.1388
0.0003	22.0	2772	0.4251	0.0075	0.1389
0.0001	23.0	2898	0.4354	0.0075	0.1361
0.0001	24.0	3024	0.4376	0.0075	0.1362

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.2