wav2vec2-xls-r-300m-MCV15

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

• Loss: 0.9816
• Wer: 0.6048
• Cer: 0.2217

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: 6e-05
• train_batch_size: 24
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 48
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 60
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
11.1167	4.5	250	3.6649	1.0	1.0000
3.1388	9.01	500	2.9534	1.0	1.0000
2.1614	13.51	750	1.3123	0.8240	0.3193
1.0783	18.02	1000	1.0311	0.7298	0.2684
0.7555	22.52	1250	0.9512	0.6806	0.2486
0.6159	27.03	1500	0.9362	0.6561	0.2418
0.5212	31.53	1750	0.9738	0.6409	0.2344
0.4684	36.04	2000	0.9576	0.6223	0.2282
0.4275	40.54	2250	0.9829	0.6178	0.2267
0.3856	45.05	2500	0.9753	0.6102	0.2244
0.3665	49.55	2750	0.9797	0.6058	0.2223
0.3668	54.05	3000	0.9690	0.6046	0.2217
0.3294	58.56	3250	0.9816	0.6048	0.2217

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.0
• Tokenizers 0.15.0