wav2vec_cv

This model is a fine-tuned version of facebook/wav2vec2-base-960h on the None dataset. It achieves the following results on the evaluation set:

• Loss: 4.1760
• Wer: 1.0

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: 0.003
• train_batch_size: 6
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 12
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 20
• num_epochs: 60

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.1467	4.29	30	4.2173	1.0
6.8918	8.57	60	4.2004	1.0
5.4913	12.86	90	4.2007	1.0
5.3906	17.14	120	4.1765	1.0
4.9212	21.43	150	4.1714	1.0
4.3916	25.71	180	4.1811	1.0
5.2255	30.0	210	4.1633	1.0
4.501	34.29	240	4.2050	1.0
4.4328	38.57	270	4.1572	1.0
4.2136	42.86	300	4.1698	1.0
4.3353	47.14	330	4.1721	1.0
4.1805	51.43	360	4.1804	1.0
4.1695	55.71	390	4.1801	1.0
4.2978	60.0	420	4.1760	1.0

Framework versions

• Transformers 4.17.0
• Pytorch 1.11.0+cu113
• Datasets 1.18.3
• Tokenizers 0.12.1