finetuning2

This model is a fine-tuned version of facebook/wav2vec2-base on the common_voice_1_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.6883
• Wer: 0.4214

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: 32
• 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: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.5277	4.27	500	2.8353	0.9863
1.2768	8.55	1000	0.7019	0.5581
0.4511	12.82	1500	0.6201	0.4726
0.2591	17.09	2000	0.6428	0.4469
0.1854	21.37	2500	0.6901	0.4388
0.1386	25.64	3000	0.6933	0.4259
0.111	29.91	3500	0.6883	0.4214

Framework versions

• Transformers 4.39.3
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.2