esm2-t36-3B-lora-16-remote-homology-filtered

This model is a fine-tuned version of facebook/esm2_t36_3B_UR50D on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4403
• Precision: 0.7922
• Recall: 0.8139
• F1-score: 0.8029
• Accuracy: 0.7990

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: 0.0002
• train_batch_size: 96
• eval_batch_size: 96
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 192
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.1
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1-score	Accuracy
0.535	0.9992	664	0.5186	0.8002	0.6630	0.7252	0.7472
0.4946	2.0	1329	0.5065	0.6945	0.8969	0.7828	0.7496
0.4727	2.9992	1993	0.4592	0.7917	0.7876	0.7897	0.7889
0.4439	4.0	2658	0.4471	0.8087	0.7798	0.7940	0.7964
0.4234	4.9962	3320	0.4403	0.7922	0.8139	0.8029	0.7990

Framework versions

• PEFT 0.10.0
• Transformers 4.40.2
• Pytorch 2.3.0+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1