Arbi-Houssem/TunLangModel_test1.15

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: 1.2386
• Wer: 62.7026

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-07
• 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
1.7779	4.5045	500	1.7860	89.0690
1.3825	9.0090	1000	1.4590	75.6419
1.2777	13.5135	1500	1.3650	72.4573
1.255	18.0180	2000	1.3039	68.0247
1.1248	22.5225	2500	1.2712	65.4139
1.0907	27.0270	3000	1.2522	62.5448
1.0969	31.5315	3500	1.2418	62.2579
1.111	36.0360	4000	1.2386	62.7026

Framework versions

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