Arbi-Houssem/TunLangModel_test1.14

Whisper Tunisien

This model is a fine-tuned version of openai/whisper-small on the Tunisian_dataset_STT-TTS15s_filtred1.0_Mixed dataset. It achieves the following results on the evaluation set:

• Loss: 0.9262
• Wer: 53.5504

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: 1e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.0933	4.5045	500	1.1956	61.0242
0.7416	9.0090	1000	0.9971	56.0895
0.5403	13.5135	1500	0.9140	53.4644
0.4315	18.0180	2000	0.8778	53.3783
0.2836	22.5225	2500	0.8948	53.0627
0.2513	27.0270	3000	0.9099	53.5074
0.2196	31.5315	3500	0.9215	53.6365
0.2199	36.0360	4000	0.9262	53.5504

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1