ap-jxLlj8Cg75EMClO0ZGP7gX

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

• Loss: 0.4326
• Model Preparation Time: 0.0053
• Wer: 0.1928

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: 0.0003
• 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: 12
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Model Preparation Time	Wer
0.8692	1.0	13	1.5918	0.0053	0.2712
0.2971	2.0	26	0.5042	0.0053	0.2033
0.1803	3.0	39	0.2987	0.0053	0.1344
0.0646	4.0	52	0.2941	0.0053	0.1340
0.0386	5.0	65	0.3093	0.0053	0.1471
0.0102	6.0	78	0.3861	0.0053	0.1438
0.0183	7.0	91	0.3980	0.0053	0.1513
0.037	8.0	104	0.4312	0.0053	0.1538
0.0164	9.0	117	0.4298	0.0053	0.1699
0.0132	10.0	130	0.3946	0.0053	0.1349
0.018	11.0	143	0.4802	0.0053	0.3464
0.0317	11.08	144	0.4326	0.0053	0.1928

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.1