ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts-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: 0.3093
- Linker Heavy Atoms Difference: 0.3372
- Has Three Substructures: 0.9991
- Poi Valid: 0.9534
- Linker Graph Edit Distance Norm: inf
- Reassembly: 0.5402
- Valid: 0.9458
- Poi Heavy Atoms Difference: 1.3755
- E3 Has Attachment Point(s): 0.9922
- Poi Heavy Atoms Difference Norm: 0.0422
- Heavy Atoms Difference: 4.5184
- Has All Attachment Points: 0.9900
- Poi Tanimoto Similarity: 0.0
- Heavy Atoms Difference Norm: 0.0596
- E3 Heavy Atoms Difference: 0.2046
- All Ligands Equal: 0.5318
- E3 Heavy Atoms Difference Norm: -0.0014
- Linker Heavy Atoms Difference Norm: 0.0044
- Tanimoto Similarity: 0.0
- E3 Graph Edit Distance Norm: inf
- Poi Equal: 0.7580
- Poi Has Attachment Point(s): 0.9534
- Poi Graph Edit Distance Norm: inf
- Num Fragments: 2.9998
- Linker Graph Edit Distance: inf
- Linker Valid: 0.9980
- Linker Tanimoto Similarity: 0.0
- Poi Graph Edit Distance: inf
- E3 Equal: 0.8017
- Linker Has Attachment Point(s): 0.9980
- Linker Equal: 0.7645
- E3 Tanimoto Similarity: 0.0
- Reassembly Nostereo: 0.5728
- E3 Valid: 0.9922
- E3 Graph Edit Distance: inf
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_with_restarts
- lr_scheduler_warmup_steps: 100
- training_steps: 10000
- mixed_precision_training: Native AMP
Training results
| Training Loss | Epoch | Step | Validation Loss | Linker Heavy Atoms Difference | Has Three Substructures | Poi Valid | Linker Graph Edit Distance Norm | Reassembly | Valid | Poi Heavy Atoms Difference | E3 Has Attachment Point(s) | Poi Heavy Atoms Difference Norm | Heavy Atoms Difference | Has All Attachment Points | Poi Tanimoto Similarity | Heavy Atoms Difference Norm | E3 Heavy Atoms Difference | All Ligands Equal | E3 Heavy Atoms Difference Norm | Linker Heavy Atoms Difference Norm | Tanimoto Similarity | E3 Graph Edit Distance Norm | Poi Equal | Poi Has Attachment Point(s) | Poi Graph Edit Distance Norm | Num Fragments | Linker Graph Edit Distance | Linker Valid | Linker Tanimoto Similarity | Poi Graph Edit Distance | E3 Equal | Linker Has Attachment Point(s) | Linker Equal | E3 Tanimoto Similarity | Reassembly Nostereo | E3 Valid | E3 Graph Edit Distance |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.0212 | 0.4932 | 5000 | 0.2852 | 0.5099 | 0.9994 | 0.9528 | inf | 0.4464 | 0.9500 | 1.4706 | 0.9979 | 0.0414 | 4.6065 | 0.9903 | 0.0 | 0.0591 | 0.1237 | 0.4396 | -0.0043 | 0.0050 | 0.0 | inf | 0.7208 | 0.9528 | inf | 3.0001 | 24787535410764872823308202666266910179524056896382338942894080.0000 | 0.9975 | 0.0 | inf | 0.7734 | 0.9975 | 0.6416 | 0.0 | 0.4783 | 0.9979 | inf |
| 0.011 | 0.7398 | 7500 | 0.3022 | 0.3255 | 0.9990 | 0.9389 | 0.0511 | 0.4896 | 0.9294 | 2.2561 | 0.9905 | 0.0673 | 6.6114 | 0.9858 | 0.0 | 0.0846 | 0.5439 | 0.4827 | 0.0117 | -0.0008 | 0.0 | inf | 0.7332 | 0.9389 | inf | 3.0001 | 27443342776203963485595838968590031296472552664861602492710912.0000 | 0.9973 | 0.0 | inf | 0.7827 | 0.9973 | 0.7075 | 0.0 | 0.5176 | 0.9905 | inf |
| 0.007 | 0.9864 | 10000 | 0.3093 | 0.3372 | 0.9991 | 0.9534 | inf | 0.5402 | 0.9458 | 1.3755 | 0.9922 | 0.0422 | 4.5184 | 0.9900 | 0.0 | 0.0596 | 0.2046 | 0.5318 | -0.0014 | 0.0044 | 0.0 | inf | 0.7580 | 0.9534 | inf | 2.9998 | inf | 0.9980 | 0.0 | inf | 0.8017 | 0.9980 | 0.7645 | 0.0 | 0.5728 | 0.9922 | inf |
Framework versions
- Transformers 4.44.2
- Pytorch 2.4.1+cu121
- Datasets 3.0.0
- Tokenizers 0.19.1
- Downloads last month
- 3
Inference Providers
NEW
This model isn't deployed by any Inference Provider.
๐
Ask for provider support
Model tree for ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts-opt25-rand-smiles
Base model
seyonec/ChemBERTa-zinc-base-v1