wav2vec-read_aloud

This model is a fine-tuned version of facebook/w2v-bert-2.0 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.1115
• Pcc Accuracy: 0.7918
• Pcc Fluency: 0.7940
• Pcc Total Score: 0.8472
• Pcc Content: 0.8160

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: 0.00055
• train_batch_size: 2
• eval_batch_size: 6
• seed: 42
• gradient_accumulation_steps: 12
• total_train_batch_size: 24
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.25
• num_epochs: 14

Training results

Training Loss	Epoch	Step	Validation Loss	Pcc Accuracy	Pcc Fluency	Pcc Total Score	Pcc Content
0.1483	1.94	500	0.1659	0.7256	0.6982	0.7616	0.7480
0.1338	3.89	1000	0.1369	0.7706	0.7680	0.8154	0.7835
0.124	5.83	1500	0.1754	0.6686	0.6459	0.7110	0.6823
0.1147	7.77	2000	0.1149	0.7838	0.7848	0.8368	0.8048
0.1024	9.72	2500	0.1135	0.7802	0.7819	0.8340	0.8048
0.0945	11.66	3000	0.1168	0.7891	0.7876	0.8418	0.8095
0.0945	13.61	3500	0.1115	0.7918	0.7940	0.8472	0.8160

Framework versions

• Transformers 4.37.0
• Pytorch 2.1.2
• Datasets 2.18.0
• Tokenizers 0.15.1