nyankole_wav2vec2-runpod-xls-r-300m-lrPlateau

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: 0.6614
• Wer: 0.5415

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-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: reduce_lr_on_plateau
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
6.1863	0.9953	105	3.1550	1.0
2.9618	2.0	211	2.8783	1.0
2.8789	2.9953	316	2.8603	1.0
2.8303	4.0	422	2.8302	1.0
2.8401	4.9953	527	2.8178	1.0
2.8001	6.0	633	2.7994	1.0
2.8123	6.9953	738	2.7633	1.0
2.6759	8.0	844	2.3806	1.0
1.9592	8.9953	949	1.3134	0.9990
1.2909	10.0	1055	0.9641	0.8247
1.0631	10.9953	1160	0.8082	0.7242
0.9105	12.0	1266	0.7254	0.6706
0.8196	12.9953	1371	0.6825	0.6552
0.7403	14.0	1477	0.6688	0.6265
0.6957	14.9953	1582	0.6459	0.5905
0.6438	16.0	1688	0.6298	0.5796
0.6147	16.9953	1793	0.6176	0.5701
0.5822	18.0	1899	0.6246	0.5562
0.5593	18.9953	2004	0.6254	0.5567
0.5334	20.0	2110	0.6184	0.5487
0.5171	20.9953	2215	0.6127	0.5459
0.475	22.0	2321	0.6093	0.5335
0.4814	22.9953	2426	0.6090	0.5485
0.4441	24.0	2532	0.5986	0.5361
0.4402	24.9953	2637	0.6142	0.5459
0.4196	26.0	2743	0.6262	0.5448
0.4109	26.9953	2848	0.6119	0.5397
0.4078	28.0	2954	0.6150	0.5412
0.3852	28.9953	3059	0.6237	0.5332
0.3731	30.0	3165	0.6301	0.5338
0.3601	30.9953	3270	0.6162	0.5461
0.3553	32.0	3376	0.6614	0.5415

Framework versions

• Transformers 4.42.4
• Pytorch 2.2.0+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1