hubert_arabic_mdd_10

This model is a fine-tuned version of facebook/hubert-large-ls960-ft on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.5464
• Wer: 0.0794
• Per: 0.0653

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Per
15.5246	1.0	164	3.2389	1.0	1.0
3.2601	2.0	328	3.1338	1.0	1.0
2.911	3.0	492	2.5069	0.9783	0.9855
2.0767	4.0	656	1.4979	0.6935	0.6882
1.2653	5.0	820	1.1179	0.4088	0.3823
0.8099	6.0	984	0.7524	0.2751	0.2546
0.5667	7.0	1148	0.6187	0.1586	0.1353
0.4083	8.0	1312	0.5950	0.1286	0.1082
0.3309	9.0	1476	0.4893	0.1137	0.0959
0.267	10.0	1640	0.5623	0.1146	0.0965
0.2349	11.0	1804	0.5520	0.0972	0.0810
0.2139	12.0	1968	0.6002	0.0939	0.0777
0.192	13.0	2132	0.5210	0.0888	0.0736
0.1559	14.0	2296	0.5566	0.0867	0.0712
0.1647	15.0	2460	0.5404	0.0886	0.0734
0.1537	16.0	2624	0.5280	0.0849	0.0702
0.1306	17.0	2788	0.5547	0.0790	0.0645
0.1269	18.0	2952	0.5435	0.0797	0.0654
0.1237	19.0	3116	0.5494	0.0802	0.0658
0.1258	20.0	3280	0.5464	0.0794	0.0653

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0