Runyankore_speech_to_intent_wav2vec_xlsr_Lg_ctc

This model is a fine-tuned version of KasuleTrevor/wav2vec2-large-xls-r-300m-lg-cv-130hr-v1 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1134
• Accuracy: 0.9836
• Precision: 0.9844
• Recall: 0.9836
• F1: 0.9837

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: 32
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 500
• num_epochs: 100
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
No log	1.0	88	2.5868	0.3360	0.2966	0.3360	0.2398
2.85	2.0	176	0.6469	0.9430	0.9596	0.9430	0.9321
1.2479	3.0	264	0.1413	0.9796	0.9807	0.9796	0.9797
0.2159	4.0	352	0.1457	0.9532	0.9681	0.9532	0.9527
0.1389	5.0	440	0.0888	0.9817	0.9828	0.9817	0.9817
0.1214	6.0	528	0.0947	0.9796	0.9806	0.9796	0.9797
0.1065	7.0	616	0.0889	0.9837	0.9848	0.9837	0.9838
0.0853	8.0	704	0.0959	0.9756	0.9788	0.9756	0.9752
0.0853	9.0	792	0.0910	0.9817	0.9826	0.9817	0.9817
0.0771	10.0	880	0.0781	0.9817	0.9831	0.9817	0.9818
0.0556	11.0	968	0.1050	0.9837	0.9850	0.9837	0.9839
0.0427	12.0	1056	0.1048	0.9817	0.9831	0.9817	0.9818
0.0454	13.0	1144	0.0870	0.9837	0.9844	0.9837	0.9837
0.0378	14.0	1232	0.1262	0.9776	0.9789	0.9776	0.9777
0.0389	15.0	1320	0.0998	0.9837	0.9848	0.9837	0.9838
0.0352	16.0	1408	0.0998	0.9796	0.9808	0.9796	0.9798
0.0352	17.0	1496	0.1053	0.9817	0.9823	0.9817	0.9817

Framework versions

• Transformers 4.51.3
• Pytorch 2.1.0+cu118
• Datasets 3.6.0
• Tokenizers 0.21.2