ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_reduce-rand-smiles-train-0.25

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: 0.3624
  • Linker Equal: 0.8351
  • E3 Graph Edit Distance Norm: inf
  • Poi Heavy Atoms Difference Norm: 0.0311
  • Poi Has Attachment Point(s): 0.9593
  • Poi Graph Edit Distance Norm: inf
  • Valid: 0.9367
  • Poi Valid: 0.9593
  • Poi Graph Edit Distance: inf
  • Reassembly Nostereo: 0.6220
  • Linker Has Attachment Point(s): 0.9970
  • Poi Heavy Atoms Difference: 1.1101
  • Has Three Substructures: 1.0
  • E3 Heavy Atoms Difference Norm: 0.0153
  • Linker Valid: 0.9970
  • Linker Tanimoto Similarity: 0.0
  • E3 Has Attachment Point(s): 0.9798
  • Poi Equal: 0.7936
  • All Ligands Equal: 0.5870
  • Heavy Atoms Difference Norm: 0.0731
  • Linker Heavy Atoms Difference Norm: 0.0030
  • Heavy Atoms Difference: 5.4708
  • Tanimoto Similarity: 0.0
  • Num Fragments: 3.0
  • E3 Graph Edit Distance: inf
  • Linker Graph Edit Distance: inf
  • Linker Heavy Atoms Difference: 0.2800
  • Poi Tanimoto Similarity: 0.0
  • Linker Graph Edit Distance Norm: inf
  • Has All Attachment Points: 0.9888
  • E3 Tanimoto Similarity: 0.0
  • Reassembly: 0.5944
  • E3 Heavy Atoms Difference: 0.4994
  • E3 Equal: 0.8276
  • E3 Valid: 0.9798

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 Linker Equal E3 Graph Edit Distance Norm Poi Heavy Atoms Difference Norm Poi Has Attachment Point(s) Poi Graph Edit Distance Norm Valid Poi Valid Poi Graph Edit Distance Reassembly Nostereo Linker Has Attachment Point(s) Poi Heavy Atoms Difference Has Three Substructures E3 Heavy Atoms Difference Norm Linker Valid Linker Tanimoto Similarity E3 Has Attachment Point(s) Poi Equal All Ligands Equal Heavy Atoms Difference Norm Linker Heavy Atoms Difference Norm Heavy Atoms Difference Tanimoto Similarity Num Fragments E3 Graph Edit Distance Linker Graph Edit Distance Linker Heavy Atoms Difference Poi Tanimoto Similarity Linker Graph Edit Distance Norm Has All Attachment Points E3 Tanimoto Similarity Reassembly E3 Heavy Atoms Difference E3 Equal E3 Valid
0.0006 31.5582 80000 0.3577 0.8324 inf 0.0305 0.9579 inf 0.9328 0.9579 inf 0.6230 0.9967 1.0626 0.9999 0.0251 0.9967 0.0 0.9771 0.7863 0.5856 0.0758 0.0051 5.7523 0.0 3.0001 inf inf 0.2647 0.0 inf 0.9890 0.0 0.5934 0.8042 0.8272 0.9771
0.0005 39.4477 100000 0.3624 0.8351 inf 0.0311 0.9593 inf 0.9367 0.9593 inf 0.6220 0.9970 1.1101 1.0 0.0153 0.9970 0.0 0.9798 0.7936 0.5870 0.0731 0.0030 5.4708 0.0 3.0 inf inf 0.2800 0.0 inf 0.9888 0.0 0.5944 0.4994 0.8276 0.9798

Framework versions

  • Transformers 4.44.2
  • Pytorch 2.4.1+cu121
  • Datasets 3.0.0
  • Tokenizers 0.19.1
Downloads last month
13
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-0.25

Finetuned
(20)
this model

Evaluation results