bert-base-cased-ChemTok-ZN15-55KTyrosinase-V1

This model is a fine-tuned version of bert-base-cased on the cafierom/ZN55K_Tyrosinase dataset of drug or drug-like molecules.

It achieves the following results on the evaluation set:

• Loss: 0.3050

Model description

This domain adaptation of bert-base-cased has been trained on ~56.5K molecular SMILES strings, with added tokens:

new_tokens = ["[C@H]","[C@@H]","(F)","(Cl)","c1","c2","(O)","N#C","(=O)",
              "([N+]([O-])=O)","[O-]","(OC)","(C)","[NH3+]","(I)","[Na+]","C#N"]

Intended uses & limitations

It is meant to be used for finetuning classification models for drug-related tasks, and for generative unmasking.

Training and evaluation data

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 128
• eval_batch_size: 128
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss
2.8365	1.0	133	1.8916
1.2975	2.0	266	0.9819
0.955	3.0	399	0.8473
0.7853	4.0	532	0.6689
0.6547	5.0	665	0.5779
0.5822	6.0	798	0.5189
0.5312	7.0	931	0.4685
0.4937	8.0	1064	0.4340
0.467	9.0	1197	0.4364
0.4442	10.0	1330	0.3819
0.4309	11.0	1463	0.4075
0.4159	12.0	1596	0.3835
0.4045	13.0	1729	0.3692
0.3874	14.0	1862	0.3672
0.3734	15.0	1995	0.3617
0.3742	16.0	2128	0.3492
0.364	17.0	2261	0.3357
0.3547	18.0	2394	0.3323
0.3435	19.0	2527	0.3222
0.3424	20.0	2660	0.3140
0.3417	21.0	2793	0.3117
0.3339	22.0	2926	0.3153
0.326	23.0	3059	0.3233
0.3292	24.0	3192	0.3019
0.324	25.0	3325	0.2960
0.3228	26.0	3458	0.3058
0.3176	27.0	3591	0.2984
0.3112	28.0	3724	0.2889
0.3172	29.0	3857	0.2973
0.3103	30.0	3990	0.3050

Framework versions

• Transformers 4.50.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1