whisper-small-ar-eos-v8-eos-v12-1.7

This model is a fine-tuned version of Ahmed107/whisper-small-ar-eos-v8 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 3.1353
• Accuracy: 0.6769

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-06
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• optimizer: Use 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_ratio: 0.2
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.6284	1.0	506	0.6770	0.5022
0.6035	2.0	1012	0.6186	0.6830
0.5059	3.0	1518	0.6554	0.6674
0.582	4.0	2024	0.5851	0.6959
0.4402	5.0	2530	0.6048	0.6975
0.3958	6.0	3036	0.6634	0.6948
0.3674	7.0	3542	0.8059	0.6948
0.1777	8.0	4048	1.0172	0.6864
0.1309	9.0	4554	1.1437	0.6747
0.0751	10.0	5060	1.5102	0.6763
0.0493	11.0	5566	1.9769	0.6680
0.0371	12.0	6072	2.3622	0.6797
0.0081	13.0	6578	2.5350	0.6802
0.0002	14.0	7084	2.8013	0.6730
0.0178	15.0	7590	2.9604	0.6663
0.0001	16.0	8096	3.0905	0.6652
0.0001	17.0	8602	3.0268	0.6669
0.0001	18.0	9108	2.8729	0.6853
0.0	19.0	9614	2.9037	0.6847
0.0171	20.0	10120	2.8267	0.6808
0.0	21.0	10626	2.8189	0.6814
0.0	22.0	11132	2.9142	0.6791
0.0001	23.0	11638	2.8557	0.6814
0.0	24.0	12144	3.0361	0.6724
0.0	25.0	12650	3.0190	0.6713
0.0	26.0	13156	3.0526	0.6752
0.0	27.0	13662	3.0976	0.6802
0.0	28.0	14168	3.1136	0.6713
0.0	29.0	14674	3.1319	0.6758
0.0	30.0	15180	3.1353	0.6769

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0