w2v2-libri-10min

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

• Loss: 2.1310
• Wer: 0.6321

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: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
5.4815	62.5	250	2.9246	1.0
2.8853	125.0	500	2.9048	1.0
1.7486	187.5	750	1.4360	0.6805
0.0923	250.0	1000	1.9166	0.6777
0.0379	312.5	1250	1.9635	0.6694
0.0209	375.0	1500	1.9195	0.6625
0.012	437.5	1750	2.1305	0.6335
0.0078	500.0	2000	2.1604	0.6169
0.0047	562.5	2250	2.1273	0.6266
0.0035	625.0	2500	2.1310	0.6321

Framework versions

• Transformers 4.29.2
• Pytorch 2.0.1+cu118
• Datasets 1.18.3
• Tokenizers 0.13.3