Ahmed107/whisper-small-ar-eos-v4

whisper-small-ar-ft-kws-speech-commands

This model is a fine-tuned version of arbml/whisper-small-ar on the Speech Commands dataset. It achieves the following results on the evaluation set:

• Loss: 3.3471
• Accuracy: 0.5748

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: 16
• eval_batch_size: 16
• 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: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.682	1.0	166	0.6867	0.6395
0.6463	2.0	332	0.6377	0.6531
0.5829	3.0	498	0.6250	0.6633
0.6197	4.0	664	0.6798	0.6429
0.3921	5.0	830	0.9584	0.5918
0.3009	6.0	996	0.9658	0.6395
0.123	7.0	1162	1.3115	0.6293
0.1418	8.0	1328	1.8621	0.6190
0.1181	9.0	1494	2.2151	0.6020
0.0014	10.0	1660	2.3968	0.6156
0.0007	11.0	1826	2.7913	0.5646
0.0004	12.0	1992	2.9198	0.6020
0.0003	13.0	2158	2.9664	0.5850
0.0002	14.0	2324	3.1507	0.5850
0.0002	15.0	2490	3.1987	0.5884
0.0001	16.0	2656	3.2650	0.5782
0.0001	17.0	2822	3.3091	0.5714
0.0002	18.0	2988	3.3048	0.5782
0.0023	19.0	3154	3.2925	0.5918
0.0001	20.0	3320	3.3471	0.5748

Framework versions

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