ailab-bio/PROTAC-Splitter-EncoderDecoder-lr_cosine_restarts

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: 0.3421
• Linker Heavy Atoms Difference Norm: 0.0015
• Poi Valid: 0.9263
• E3 Graph Edit Distance: inf
• Poi Graph Edit Distance: 737429178470255015019840581324076533822563653950209220564484096.0000
• Poi Has Attachment Point(s): 0.9263
• E3 Has Attachment Point(s): 0.9785
• Poi Heavy Atoms Difference Norm: 0.0735
• Poi Equal: 0.7917
• Linker Tanimoto Similarity: 0.0
• Linker Graph Edit Distance: inf
• Linker Equal: 0.8455
• Linker Valid: 0.9975
• Reassembly: 0.6035
• Has All Attachment Points: 0.9868
• Heavy Atoms Difference Norm: 0.1010
• Reassembly Nostereo: 0.6258
• E3 Graph Edit Distance Norm: inf
• All Ligands Equal: 0.5964
• E3 Equal: 0.8260
• E3 Valid: 0.9785
• E3 Heavy Atoms Difference Norm: 0.0174
• Heavy Atoms Difference: 7.5459
• E3 Tanimoto Similarity: 0.0
• Has Three Substructures: 0.9998
• E3 Heavy Atoms Difference: 0.6771
• Linker Heavy Atoms Difference: 0.2282
• Linker Graph Edit Distance Norm: inf
• Linker Has Attachment Point(s): 0.9975
• Valid: 0.9041
• Poi Tanimoto Similarity: 0.0
• Poi Heavy Atoms Difference: 2.1196
• Tanimoto Similarity: 0.0
• Num Fragments: 3.0002
• Poi Graph Edit Distance Norm: 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: 800
• training_steps: 100000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	All Ligands Equal	E3 Equal	E3 Graph Edit Distance	E3 Graph Edit Distance Norm	E3 Has Attachment Point(s)	E3 Heavy Atoms Difference	E3 Heavy Atoms Difference Norm	E3 Tanimoto Similarity	E3 Valid	Has All Attachment Points	Has Three Substructures	Heavy Atoms Difference	Heavy Atoms Difference Norm	Linker Equal	Linker Graph Edit Distance	Linker Graph Edit Distance Norm	Linker Has Attachment Point(s)	Linker Heavy Atoms Difference	Linker Heavy Atoms Difference Norm	Linker Tanimoto Similarity	Linker Valid	Validation Loss	Num Fragments	Poi Equal	Poi Graph Edit Distance	Poi Graph Edit Distance Norm	Poi Has Attachment Point(s)	Poi Heavy Atoms Difference	Poi Heavy Atoms Difference Norm	Poi Tanimoto Similarity	Poi Valid	Reassembly	Reassembly Nostereo	Tanimoto Similarity	Valid
0.0156	0.4932	5000	0.4596	0.7815	inf	inf	0.9916	0.3913	0.0061	0.0	0.9916	0.9863	0.9988	7.0030	0.0918	0.6623	35410764872521246890440289523238443113605795566260484204134400.0000	inf	0.9965	0.3805	0.0005	0.0	0.9965	0.2891	3.0006	0.7265	inf	inf	0.9252	2.3129	0.0739	0.0	0.9252	0.4665	0.4911	0.0	0.9169
0.0077	0.7398	7500	0.4898	0.7908	inf	inf	0.9857	0.4845	0.0088	0.0	0.9857	0.9812	0.9995	8.2839	0.1093	0.7095	56657223796033995024704463237181508981769272906016774726615040.0000	inf	0.9943	0.4483	0.0090	0.0	0.9943	0.3073	3.0002	0.7363	inf	inf	0.9161	2.5750	0.0832	0.0	0.9161	0.4975	0.5220	0.0	0.8987
0.0046	0.9864	10000	0.5462	0.8045	inf	inf	0.9896	0.5553	0.0131	0.0	0.9896	0.9836	0.9991	7.0102	0.0939	0.7666	inf	inf	0.9961	0.3144	0.0033	0.0	0.9961	0.3086	2.9998	0.7680	inf	inf	0.9272	2.1208	0.0719	0.0	0.9272	0.5544	0.5796	0.0	0.9157
0.0003	6.9047	70000	0.3416	0.0013	0.9247	inf	inf	0.9247	0.9850	0.0743	0.7900	0.0	28328611898017003221343002442430278724028514250988932894294016.0000	0.8408	0.9972	0.5992	0.9859	0.0987	0.6233	inf	0.5914	0.8266	0.9850	0.0135	7.3665	0.0	1.0	0.5428	0.2086	inf	0.9972	0.9082	0.0	2.1749	0.0	3.0	inf
0.0004	7.1513	72500	0.3364	0.0027	0.9271	inf	inf	0.9271	0.9849	0.0707	0.7934	0.0	inf	0.8451	0.9971	0.6019	0.9846	0.0960	0.6242	inf	0.5942	0.8275	0.9849	0.0122	7.1871	0.0	1.0	0.5763	0.2224	inf	0.9971	0.9109	0.0	2.0791	0.0	3.0	inf
0.0005	7.3979	75000	0.3489	0.0030	0.9274	inf	inf	0.9274	0.9839	0.0704	0.7870	0.0	inf	0.8362	0.9971	0.5971	0.9873	0.0985	0.6222	inf	0.5900	0.8249	0.9839	0.0187	7.4559	0.0	0.9997	0.6882	0.2505	inf	0.9971	0.9108	0.0	2.1431	0.0	2.9997	inf
0.0012	7.6445	77500	0.3375	0.0072	0.9271	inf	inf	0.9271	0.9820	0.0760	0.7898	0.0	inf	0.8247	0.9965	0.5921	0.9864	0.0983	0.6127	inf	0.5848	0.8250	0.9820	0.0149	7.4127	0.0	0.9997	0.6424	0.3465	inf	0.9965	0.9095	0.0	2.2373	0.0	2.9997	inf
0.0003	7.8911	80000	0.3411	0.0019	0.9295	inf	inf	0.9295	0.9783	0.0750	0.7928	0.0	inf	0.8451	0.9971	0.6030	0.9868	0.0997	0.6278	inf	0.5955	0.8263	0.9783	0.0179	7.4869	0.0	0.9995	0.7170	0.2144	inf	0.9971	0.9079	0.0	2.1623	0.0	2.9998	inf
0.0004	8.1377	82500	0.3505	0.0009	0.9268	inf	inf	0.9268	0.9796	0.0742	0.7927	0.0	inf	0.8377	0.9966	0.5997	0.9863	0.1033	0.6243	inf	0.5924	0.8275	0.9796	0.0176	7.7488	0.0	0.9999	0.6833	0.2141	inf	0.9966	0.9047	0.0	2.1771	0.0	3.0001	inf
0.0005	8.3843	85000	0.3462	0.0021	0.9173	inf	inf	0.9173	0.9818	0.0824	0.7883	0.0	inf	0.8387	0.9964	0.5984	0.9820	0.1082	0.6215	inf	0.5915	0.8260	0.9818	0.0158	8.1167	0.0	0.9996	0.6591	0.2392	inf	0.9964	0.8986	0.0	2.4100	0.0	3.0	inf
0.0009	8.6309	87500	0.3325	0.0020	0.9225	inf	inf	0.9225	0.9796	0.0794	0.7900	0.0	inf	0.8345	0.9972	0.5966	0.9834	0.1044	0.6196	inf	0.5887	0.8239	0.9796	0.0165	7.8794	0.0	0.9996	0.6269	0.2329	inf	0.9972	0.9019	0.0	2.3541	0.0	2.9996	inf
0.0003	8.8775	90000	0.3373	0.0020	0.9258	inf	inf	0.9258	0.9801	0.0744	0.7946	0.0	25672804532577906850064595316267633373936140684529123813490688.0000	0.8446	0.9974	0.6029	0.9847	0.1014	0.6285	inf	0.5953	0.8287	0.9801	0.0175	7.6187	0.0	0.9999	0.6809	0.2217	inf	0.9974	0.9058	0.0	2.1887	0.0	2.9999	inf
0.0003	9.1241	92500	0.3412	0.0040	0.9302	inf	inf	0.9302	0.9817	0.0684	0.7867	0.0	31869688385269122201396260570914598802245216009634435783720960.0000	0.8426	0.9968	0.6011	0.9828	0.0971	0.6213	inf	0.5941	0.8295	0.9817	0.0157	7.3307	0.0	0.9997	0.6020	0.2689	inf	0.9968	0.9106	0.0	2.0255	0.0	2.9997	inf
0.0004	9.3707	95000	0.3420	0.0071	0.9275	inf	inf	0.9275	0.9796	0.0735	0.7880	0.0	23902266288951841651047195428185949101683912007225826837790720.0000	0.8352	0.9976	0.5992	0.9868	0.1005	0.6217	inf	0.5911	0.8257	0.9796	0.0189	7.5129	0.0	0.9996	0.6888	0.3183	inf	0.9976	0.9073	0.0	2.1638	0.0	2.9998	inf
0.001	9.6173	97500	0.3328	0.0020	0.9207	inf	inf	0.9207	0.9771	0.0780	0.7890	0.0	35410764872521246890440289523238443113605795566260484204134400.0000	0.8340	0.9965	0.5977	0.9896	0.1080	0.6202	inf	0.5900	0.8252	0.9771	0.0217	8.1480	0.0	0.9993	0.7837	0.2400	inf	0.9965	0.8982	0.0	2.2722	0.0	3.0004	inf
0.0003	9.8639	100000	0.3421	0.0015	0.9263	inf	737429178470255015019840581324076533822563653950209220564484096.0000	0.9263	0.9785	0.0735	0.7917	0.0	inf	0.8455	0.9975	0.6035	0.9868	0.1010	0.6258	inf	0.5964	0.8260	0.9785	0.0174	7.5459	0.0	0.9998	0.6771	0.2282	inf	0.9975	0.9041	0.0	2.1196	0.0	3.0002	inf

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.0
• Tokenizers 0.19.1