whisper-mmr-embeds-10h

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: 0.4848
• Wer: 19.1566

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: 9
• num_epochs: 3.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.7998	0.2727	9	1.1718	36.9511
0.6661	0.5455	18	1.0265	36.7095
0.5443	0.8182	27	0.8771	32.9587
0.4818	1.0909	36	0.7319	27.1413
0.4164	1.3636	45	0.6286	26.3361
0.443	1.6364	54	0.5728	22.9778
0.3672	1.9091	63	0.5268	22.0049
0.3724	2.1818	72	0.5015	20.5690
0.3562	2.4545	81	0.4908	20.3509
0.3661	2.7273	90	0.4861	19.1163
0.3645	3.0	99	0.4848	19.1566

Framework versions

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