sh_sr_model

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: 2.0234
• Wer: 0.4765
• Cer: 0.8746

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: 1e-05
• 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: 100
• training_steps: 1000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.9917	20.0	100	1.3445	0.5882	0.8796
0.5644	40.0	200	1.3491	0.4941	0.875
0.3946	60.0	300	1.7289	0.5412	0.8762
0.2667	80.0	400	1.8795	0.5235	0.8762
0.2559	100.0	500	2.0205	0.5235	0.8772
0.2148	120.0	600	1.8615	0.4941	0.875
0.1694	140.0	700	1.9697	0.4765	0.8746
0.1793	160.0	800	1.9240	0.4706	0.8732
0.1598	180.0	900	2.0063	0.4765	0.8742
0.1569	200.0	1000	2.0234	0.4765	0.8746

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0