Wav2Vec2Phoneme-CSfinetune

This model is a fine-tuned version of facebook/wav2vec2-lv-60-espeak-cv-ft on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 1.7640
• Wer: 0.3064
• Cer: 0.2173

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 14
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
2.7571	2.2805	500	3.4129	0.4619	0.3238
2.0486	4.5610	1000	1.9384	0.3985	0.2789
1.7859	6.8415	1500	2.9810	0.3588	0.2583
1.505	9.1220	2000	2.2573	0.3257	0.2305
1.3879	11.4025	2500	1.7517	0.3109	0.2206
1.3334	13.6830	3000	1.7640	0.3064	0.2173

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0
• Datasets 2.19.2
• Tokenizers 0.19.1