whisper-small-ar-eos-v8-eos-v12-2.5

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: 3.3173
• Accuracy: 0.6657

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
0.5805	1.0	506	0.7427	0.5
0.5596	2.0	1012	0.6658	0.6306
0.4595	3.0	1518	0.7195	0.6217
0.5451	4.0	2024	0.6221	0.6791
0.4334	5.0	2530	0.6253	0.6936
0.3446	6.0	3036	0.7335	0.6730
0.3106	7.0	3542	0.8420	0.6702
0.1988	8.0	4048	1.0634	0.6842
0.1554	9.0	4554	1.2181	0.6747
0.0698	10.0	5060	1.5398	0.6802
0.1746	11.0	5566	1.8350	0.6652
0.0639	12.0	6072	2.3253	0.6602
0.0004	13.0	6578	2.4828	0.6663
0.0338	14.0	7084	2.5842	0.6758
0.0006	15.0	7590	2.7395	0.6713
0.0002	16.0	8096	2.7720	0.6669
0.0001	17.0	8602	2.7865	0.6685
0.0002	18.0	9108	2.8497	0.6607
0.0002	19.0	9614	2.8792	0.6607
0.008	20.0	10120	2.9188	0.6691
0.0001	21.0	10626	2.7606	0.6920
0.0051	22.0	11132	3.0888	0.6669
0.0	23.0	11638	3.0709	0.6747
0.0	24.0	12144	3.1538	0.6730
0.0	25.0	12650	3.2574	0.6646
0.0	26.0	13156	3.2662	0.6663
0.0	27.0	13662	3.2635	0.6685
0.0	28.0	14168	3.2895	0.6669
0.0	29.0	14674	3.3094	0.6652
0.0	30.0	15180	3.3173	0.6657

Framework versions

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