esm2-baseline-gfp

This model is a fine-tuned version of facebook/esm2_t12_35M_UR50D on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0490
• Spearman: 0.8315
• Pearson: 0.9644
• Mse: 0.0490

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: 5e-05
• train_batch_size: 32
• eval_batch_size: 64
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 671
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Spearman	Pearson	Mse
0.6191	1.0	671	0.4987	0.4570	0.5376	0.4987
0.2340	2.0	1342	0.3202	0.6402	0.7796	0.3202
0.3390	3.0	2013	0.1728	0.7054	0.8709	0.1728
0.2688	4.0	2684	0.1814	0.7494	0.8736	0.1814
0.2030	5.0	3355	0.2359	0.4736	0.8341	0.2359
0.1101	6.0	4026	0.1292	0.7856	0.9031	0.1292
0.1786	7.0	4697	0.0888	0.7964	0.9374	0.0888
0.0878	8.0	5368	0.2250	0.7921	0.9124	0.2250
0.0690	9.0	6039	0.0967	0.8045	0.9408	0.0967
0.0774	10.0	6710	0.1106	0.8103	0.9351	0.1106
0.0573	11.0	7381	0.0727	0.8150	0.9484	0.0727
0.0443	12.0	8052	0.0690	0.8214	0.9524	0.0690
0.0503	13.0	8723	0.0615	0.8202	0.9565	0.0615
0.0233	14.0	9394	0.0522	0.8259	0.9620	0.0522
0.0229	15.0	10065	0.0607	0.8263	0.9576	0.0607
0.0250	16.0	10736	0.0513	0.8279	0.9632	0.0513
0.0285	17.0	11407	0.0521	0.8286	0.9624	0.0521
0.0265	18.0	12078	0.0507	0.8308	0.9634	0.0507
0.0163	19.0	12749	0.0494	0.8313	0.9642	0.0494
0.0382	20.0	13420	0.0490	0.8315	0.9644	0.0490

Framework versions

• Transformers 5.6.2
• Pytorch 2.11.0+cu130
• Datasets 4.8.5
• Tokenizers 0.22.2