whisper-20hrs-random-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: 0.3460
• Wer: 16.4961

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: 43
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.758	0.2986	43	0.6442	29.1173
0.4139	0.5972	86	0.4089	19.8812
0.4268	0.8958	129	0.3813	17.8817
0.3995	1.1944	172	0.3673	18.6936
0.3855	1.4931	215	0.3587	17.2409
0.3669	1.7917	258	0.3537	16.3586
0.3253	2.0903	301	0.3495	16.2915
0.3231	2.3889	344	0.3482	16.3250
0.3313	2.6875	387	0.3468	16.4190
0.3158	2.9861	430	0.3460	16.4961

Framework versions

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