wav2vec2-large-xls-r-300m-lg

This model is a fine-tuned version of Alvin-Nahabwe/wav2vec2-large-xls-r-300m-gn on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2283
• Wer: 0.1569

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: 0.0003
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 22
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.2883	1.39	400	0.2180	0.2128
0.2603	2.78	800	0.2108	0.2082
0.2341	4.17	1200	0.2127	0.2085
0.2194	5.56	1600	0.2004	0.2050
0.1846	6.95	2000	0.1961	0.1898
0.1627	8.34	2400	0.1919	0.1779
0.1464	9.75	2800	0.1867	0.1677
0.1273	11.14	3200	0.1949	0.1710
0.1153	12.53	3600	0.1965	0.1639
0.1006	13.93	4000	0.1983	0.1603
0.1056	15.32	4400	0.2159	0.1686
0.1011	16.71	4800	0.2104	0.1663
0.0895	18.1	5200	0.2211	0.1634
0.0818	19.49	5600	0.2234	0.1610
0.0778	20.88	6000	0.2283	0.1569

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3