whisper-20hrs-meta-seed42

This model is a fine-tuned version of openai/whisper-large-v2 on the JASMIN-CGN dataset. It achieves the following results on the evaluation set:

• Loss: 0.3268
• Wer: 16.4290

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: 0.0001
• train_batch_size: 48
• eval_batch_size: 32
• seed: 42
• 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: 49
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.6911	0.2988	49	0.5748	26.2556
0.3799	0.5976	98	0.3924	17.9454
0.3565	0.8963	147	0.3631	16.9121
0.3484	1.1951	196	0.3498	15.8788
0.3296	1.4939	245	0.3406	15.7077
0.3305	1.7927	294	0.3350	16.4961
0.3278	2.0915	343	0.3311	16.5766
0.3745	2.3902	392	0.3283	16.4290
0.3125	2.6890	441	0.3271	16.2210
0.2794	2.9878	490	0.3268	16.4290

Framework versions

• PEFT 0.16.0
• Transformers 4.57.3
• Pytorch 2.7.1+cu126
• Datasets 3.6.0
• Tokenizers 0.22.2