BERT_prediction-6_tokenized

This model is a fine-tuned version of armheb/DNA_bert_6 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.0326

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: 2e-05
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.0646	1.0	276	0.0370
0.0386	2.0	552	0.0348
0.0358	3.0	828	0.0336
0.0363	4.0	1104	0.0333
0.0357	5.0	1380	0.0334
0.0341	6.0	1656	0.0342
0.0342	7.0	1932	0.0341
0.0339	8.0	2208	0.0333
0.0332	9.0	2484	0.0344
0.0345	10.0	2760	0.0323

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.1+cu113
• Datasets 2.6.1
• Tokenizers 0.13.1