esm2_t6_8M_UR50D-Ab-CLIP-v0

This model is a fine-tuned version of facebook/esm2_t6_8M_UR50D on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.8743
• Alignment: 1.0213
• Uniformity: -3.0773
• Contrastive Accuracy: 0.2267
• Top 3 Accuracy: 0.5067

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.98) and epsilon=1e-06
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 5
• num_epochs: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Alignment	Uniformity	Contrastive Accuracy	Top 3 Accuracy
1.0153	1.0	196	1.2958	1.2729	-2.4114	0.0680	0.1837
0.4623	2.0	392	1.0729	1.1755	-2.8008	0.0952	0.2653
0.2646	3.0	588	1.0867	1.1511	-2.9630	0.1361	0.3265
0.1815	4.0	784	1.0768	1.1490	-3.0403	0.1293	0.3673
0.1403	5.0	980	1.0527	1.1426	-3.0663	0.1224	0.3605

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.0+cu121
• Tokenizers 0.19.1