wav2vec_arabic_mdd_experiment2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2053
• Per: 0.1078

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: 0.0005
• train_batch_size: 8
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Per
6.6285	1.0	102	2.6113	1.0
2.3387	2.0	205	2.3622	1.0
2.3181	3.0	307	2.3030	0.9825
2.1968	4.0	410	1.9821	0.9555
1.7187	5.0	512	1.6775	0.9242
1.6233	6.0	615	1.6232	0.9440
1.5713	7.0	717	1.5419	0.9139
1.4064	8.0	820	1.3078	0.8956
1.2103	9.0	922	1.1363	0.8763
0.898	10.0	1025	0.8109	0.8318
0.6426	11.0	1127	0.6708	0.5635
0.4305	12.0	1230	0.4225	0.2889
0.2961	13.0	1332	0.3487	0.1892
0.1856	14.0	1435	0.2493	0.1223
0.137	15.0	1537	0.2439	0.1198
0.1115	16.0	1640	0.2321	0.1196
0.0923	17.0	1742	0.2226	0.1145
0.0796	18.0	1845	0.2115	0.1120
0.0742	19.0	1947	0.2082	0.1078
0.0684	19.9	2040	0.2053	0.1078

Framework versions

• Transformers 4.35.0
• Pytorch 2.1.0+cu118
• Datasets 2.14.6
• Tokenizers 0.14.1