nyankole_wav2vec2-runpod-unf-large

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 830.9091
• Wer: 0.6101

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: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7399.694	1.0	53	3616.8877	1.0
7452.8927	2.0	106	3606.0696	1.0
7321.2547	3.0	159	3614.7188	1.0
7283.513	4.0	212	3665.7957	1.0
7260.0973	5.0	265	3556.3738	1.0
7184.4652	6.0	318	3440.7732	1.0
6526.0413	7.0	371	2693.0374	1.0
5025.6185	8.0	424	1748.6117	0.9985
3761.0044	9.0	477	1390.0330	0.8946
3265.3567	10.0	530	1185.7458	0.8108
2818.3317	11.0	583	1076.5151	0.7582
2666.5274	12.0	636	1016.1800	0.7206
2484.5062	13.0	689	1017.9598	0.7026
2363.0265	14.0	742	935.6443	0.6879
2215.8507	15.0	795	926.4963	0.6670
2169.1753	16.0	848	892.7558	0.6608
2054.9318	17.0	901	885.1797	0.6461
2001.906	18.0	954	869.5634	0.6369
2011.2323	19.0	1007	857.0482	0.6317
1900.8241	20.0	1060	855.6694	0.625
1869.7319	21.0	1113	842.1160	0.6296
1809.5601	22.0	1166	838.9195	0.6219
1828.7354	23.0	1219	830.1293	0.6198
1804.8732	24.0	1272	835.9683	0.6173
1775.2409	25.0	1325	831.1539	0.6103
1805.726	26.0	1378	831.0961	0.6157
1738.365	27.0	1431	831.4120	0.6126
1802.3348	28.0	1484	831.5934	0.6113
1806.6047	29.0	1537	831.4467	0.6101
1788.635	30.0	1590	830.9091	0.6101

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1