whisper-10hrs-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: 0.3582
• Wer: 16.6236

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: 25
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.831	0.3012	25	0.9551	34.9482
0.4428	0.6024	50	0.4777	19.8410
0.425	0.9036	75	0.4106	19.3478
0.3799	1.2048	100	0.3885	18.2138
0.4097	1.5060	125	0.3755	18.0293
0.3131	1.8072	150	0.3669	17.0430
0.3489	2.1084	175	0.3634	17.2980
0.332	2.4096	200	0.3612	17.2241
0.3392	2.7108	225	0.3582	16.6236

Framework versions

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