CV11_finetuning1

This model is a fine-tuned version of Roshana/Wav2Vec1_CV on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7162
• Wer: 0.3625

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.5067	0.86	400	0.6193	0.4492
0.4448	1.72	800	0.6325	0.4384
0.3781	2.59	1200	0.6248	0.4197
0.3172	3.45	1600	0.6408	0.4343
0.2556	4.31	2000	0.6593	0.4230
0.2148	5.17	2400	0.6742	0.3987
0.1779	6.03	2800	0.6658	0.3929
0.1446	6.9	3200	0.6768	0.3846
0.1248	7.76	3600	0.6809	0.3804
0.108	8.62	4000	0.7214	0.3683
0.0938	9.48	4400	0.7162	0.3625

Framework versions

• Transformers 4.22.2
• Pytorch 1.12.1+cu113
• Datasets 2.5.1
• Tokenizers 0.12.1