longpause-300m-1109-1

This model is a fine-tuned version of SiRoZaRuPa/300m-0828-3 on the audiofolder dataset. It achieves the following results on the evaluation set:

• Loss: 4.6617
• Cer: 0.9984

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-06
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 1500
• num_epochs: 500

Training results

Training Loss	Epoch	Step	Validation Loss	Cer
30.1798	47.62	500	27.0098	0.9984
8.478	95.24	1000	8.2299	0.9984
6.247	142.86	1500	6.3422	0.9984
5.5716	190.48	2000	5.6596	0.9984
5.1403	238.1	2500	5.2172	0.9984
4.8672	285.71	3000	4.9380	0.9984
4.7082	333.33	3500	4.7765	0.9984
4.6271	380.95	4000	4.6959	0.9984
4.5957	428.57	4500	4.6659	0.9984
4.5896	476.19	5000	4.6617	0.9984

Framework versions

• Transformers 4.32.1
• Pytorch 2.0.1
• Datasets 2.12.0
• Tokenizers 0.13.2