MathildeB3/52Hz-small-LORA-fr

52Hz Small Fr - IMT Atlantique X 52 Hertz

This model is a fine-tuned version of openai/whisper-small on the Premier dataset organisé de 52 Hertz dataset. It achieves the following results on the evaluation set:

• Loss: 0.4950
• Wer: 40.3361

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.381	1.0	23	1.3088	118.0672
0.9219	2.0	46	0.8130	87.9552
0.5428	3.0	69	0.6170	44.9580
0.3826	4.0	92	0.5290	40.7563
0.2591	5.0	115	0.4884	41.4566
0.1828	6.0	138	0.4653	41.7367
0.1448	7.0	161	0.4822	40.7563
0.1124	8.0	184	0.4758	39.6359
0.0862	9.0	207	0.4789	39.6359
0.074	10.0	230	0.4870	40.1961
0.0463	11.0	253	0.4877	39.4958
0.0481	12.0	276	0.4899	39.9160
0.0346	13.0	299	0.4945	40.7563
0.041	14.0	322	0.4953	40.6162
0.034	15.0	345	0.4950	40.3361

Framework versions

• PEFT 0.18.1
• Transformers 4.57.3
• Pytorch 2.9.1+cu130
• Datasets 4.4.2
• Tokenizers 0.22.2