ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine-opt25-rand-smiles

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: 0.3184
  • E3 Tanimoto Similarity: 0.0
  • Poi Equal: 0.7620
  • E3 Equal: 0.8036
  • Poi Tanimoto Similarity: 0.0
  • Poi Valid: 0.9589
  • Reassembly: 0.5459
  • All Ligands Equal: 0.5390
  • Poi Graph Edit Distance: inf
  • Linker Graph Edit Distance: 28328611898016997512352231618590754490884636453008387363307520.0000
  • Poi Graph Edit Distance Norm: inf
  • Linker Tanimoto Similarity: 0.0
  • Valid: 0.9547
  • E3 Graph Edit Distance Norm: inf
  • Poi Has Attachment Point(s): 0.9589
  • Linker Equal: 0.7668
  • Linker Heavy Atoms Difference Norm: 0.0034
  • Has All Attachment Points: 0.9905
  • Heavy Atoms Difference: 4.1049
  • Has Three Substructures: 0.9992
  • Tanimoto Similarity: 0.0
  • Heavy Atoms Difference Norm: 0.0536
  • Poi Heavy Atoms Difference Norm: 0.0352
  • Linker Has Attachment Point(s): 0.9972
  • E3 Heavy Atoms Difference: 0.2976
  • E3 Graph Edit Distance: inf
  • Linker Valid: 0.9972
  • E3 Heavy Atoms Difference Norm: 0.0033
  • E3 Has Attachment Point(s): 0.9966
  • E3 Valid: 0.9966
  • Reassembly Nostereo: 0.5799
  • Linker Heavy Atoms Difference: 0.3031
  • Num Fragments: 3.0003
  • Linker Graph Edit Distance Norm: inf
  • Poi Heavy Atoms Difference: 1.1856

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: cosine
  • lr_scheduler_warmup_steps: 699
  • training_steps: 10000
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss E3 Tanimoto Similarity Poi Equal E3 Equal Poi Tanimoto Similarity Poi Valid Reassembly All Ligands Equal Poi Graph Edit Distance Linker Graph Edit Distance Poi Graph Edit Distance Norm Linker Tanimoto Similarity Valid E3 Graph Edit Distance Norm Poi Has Attachment Point(s) Linker Equal Linker Heavy Atoms Difference Norm Has All Attachment Points Heavy Atoms Difference Has Three Substructures Tanimoto Similarity Heavy Atoms Difference Norm Poi Heavy Atoms Difference Norm Linker Has Attachment Point(s) E3 Heavy Atoms Difference E3 Graph Edit Distance Linker Valid E3 Heavy Atoms Difference Norm E3 Has Attachment Point(s) E3 Valid Reassembly Nostereo Linker Heavy Atoms Difference Num Fragments Linker Graph Edit Distance Norm Poi Heavy Atoms Difference
0.012 0.4932 5000 0.3055 0.0 0.7354 0.7897 0.0 0.9479 0.4899 0.4824 inf 25672804532577903995569209904347871257364201785538851047997440.0000 inf 0.0 0.9441 inf 0.9479 0.6946 0.0103 0.9873 4.9586 0.9995 0.0 0.0650 0.0421 0.9974 0.1648 inf 0.9974 -0.0036 0.9972 0.9972 0.5198 0.5344 3.0005 inf 1.4326
0.0064 0.7398 7500 0.3156 0.0 0.7551 0.8017 0.0 0.9532 0.5359 0.5282 inf 40722379603399433924006332951724209580646664901199556834754560.0000 inf 0.0 0.9477 inf 0.9532 0.7533 -0.0000 0.9873 4.5195 0.9985 0.0 0.0593 0.0396 0.9959 0.3429 inf 0.9959 0.0053 0.9944 0.9944 0.5689 0.2564 3.0003 0.0452 1.3395
0.0051 0.9864 10000 0.3184 0.0 0.7620 0.8036 0.0 0.9589 0.5459 0.5390 inf 28328611898016997512352231618590754490884636453008387363307520.0000 inf 0.0 0.9547 inf 0.9589 0.7668 0.0034 0.9905 4.1049 0.9992 0.0 0.0536 0.0352 0.9972 0.2976 inf 0.9972 0.0033 0.9966 0.9966 0.5799 0.3031 3.0003 inf 1.1856

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.4.1+cu121
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
2
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_cosine-opt25-rand-smiles

Finetuned
(20)
this model

Evaluation results