ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_reduce-rand-smiles-train-0.75

This model is a fine-tuned version of seyonec/ChemBERTa-zinc-base-v1 on the ailab-bio/PROTAC-Splitter-Dataset dataset. It achieves the following results on the evaluation set:

  • Loss: 0.3562
  • E3 Heavy Atoms Difference: 0.5450
  • Tanimoto Similarity: 0.0
  • Linker Heavy Atoms Difference: 0.2341
  • Poi Graph Edit Distance: inf
  • Poi Heavy Atoms Difference Norm: 0.0500
  • Reassembly Nostereo: 0.6352
  • Has All Attachment Points: 0.9914
  • Num Fragments: 2.9996
  • Linker Equal: 0.8443
  • Reassembly: 0.6021
  • E3 Graph Edit Distance Norm: inf
  • Poi Valid: 0.9499
  • E3 Graph Edit Distance: inf
  • E3 Equal: 0.8238
  • Poi Heavy Atoms Difference: 1.5246
  • E3 Heavy Atoms Difference Norm: 0.0139
  • Has Three Substructures: 0.9996
  • Linker Valid: 0.9984
  • Linker Has Attachment Point(s): 0.9984
  • Poi Graph Edit Distance Norm: inf
  • Heavy Atoms Difference Norm: 0.0733
  • Valid: 0.9362
  • E3 Valid: 0.9859
  • Linker Heavy Atoms Difference Norm: 0.0034
  • Poi Tanimoto Similarity: 0.0
  • Poi Equal: 0.7947
  • Linker Tanimoto Similarity: 0.0
  • All Ligands Equal: 0.5946
  • Linker Graph Edit Distance Norm: inf
  • Poi Has Attachment Point(s): 0.9499
  • E3 Has Attachment Point(s): 0.9859
  • Linker Graph Edit Distance: 15934844192634561100698130285457299401122608004817217891860480.0000
  • E3 Tanimoto Similarity: 0.0
  • Heavy Atoms Difference: 5.5146

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: 128
  • eval_batch_size: 64
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: reduce_lr_on_plateau
  • lr_scheduler_warmup_steps: 400
  • training_steps: 100000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss E3 Heavy Atoms Difference Tanimoto Similarity Linker Heavy Atoms Difference Poi Graph Edit Distance Poi Heavy Atoms Difference Norm Reassembly Nostereo Has All Attachment Points Num Fragments Linker Equal Reassembly E3 Graph Edit Distance Norm Poi Valid E3 Graph Edit Distance E3 Equal Poi Heavy Atoms Difference E3 Heavy Atoms Difference Norm Has Three Substructures Linker Valid Linker Has Attachment Point(s) Poi Graph Edit Distance Norm Heavy Atoms Difference Norm Valid E3 Valid Linker Heavy Atoms Difference Norm Poi Tanimoto Similarity Poi Equal Linker Tanimoto Similarity All Ligands Equal Linker Graph Edit Distance Norm Poi Has Attachment Point(s) E3 Has Attachment Point(s) Linker Graph Edit Distance E3 Tanimoto Similarity Heavy Atoms Difference
0.0006 10.5208 80000 0.3546 0.2397 0.0 0.2432 inf 0.0551 0.6277 0.9911 2.9999 0.8418 0.5965 inf 0.9407 inf 0.8234 1.7421 0.0025 0.9999 0.9980 0.9980 inf 0.0768 0.9301 0.9907 0.0018 0.0 0.7882 0.0 0.5897 inf 0.9407 0.9907 20361189801699711253012395652022580557179454652619232886390784.0000 0.0 5.7380
0.0006 13.1510 100000 0.3562 0.5450 0.0 0.2341 inf 0.0500 0.6352 0.9914 2.9996 0.8443 0.6021 inf 0.9499 inf 0.8238 1.5246 0.0139 0.9996 0.9984 0.9984 inf 0.0733 0.9362 0.9859 0.0034 0.0 0.7947 0.0 0.5946 inf 0.9499 0.9859 15934844192634561100698130285457299401122608004817217891860480.0000 0.0 5.5146

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.4.1+cu121
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
9
Safetensors
Model size
0.1B params
Tensor type
F32
ยท
Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_reduce-rand-smiles-train-0.75

Finetuned
(20)
this model

Evaluation results