bert-base-cased-finetuned-MAOB-IC50s-V1

This model is a fine-tuned version of bert-base-cased on 4828 MAOB IC50 values from ChEMBL. It achieves the following results on the evaluation set:

• Loss: 0.7970
• Accuracy: 0.7476
• F1: 0.7478

Use in Pipeline:

from transformers import pipeline

ic50_pipe = pipeline("text-classification", model="cafierom/bert-base-cased-finetuned-MAOB-IC50s-V1")

bert_ic50 = ic50_pipe('C#CCN(C)[C@H](C)Cc1ccccc1')

Model description

More information needed

Intended uses & limitations

Can classify MAOB IC50 values as < 50 nM, < 500 nM, and > 500 nM. See Confusion matrix below:

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• 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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1
0.8701	1.0	129	0.7877	0.6731	0.5868
0.7853	2.0	258	0.8168	0.6690	0.5996
0.713	3.0	387	0.7359	0.7090	0.6881
0.6641	4.0	516	0.7053	0.6966	0.6739
0.612	5.0	645	0.7102	0.7159	0.7054
0.5773	6.0	774	0.6853	0.7283	0.7132
0.5425	7.0	903	0.6966	0.7283	0.7301
0.5085	8.0	1032	0.7151	0.7297	0.7240
0.4813	9.0	1161	0.7105	0.7379	0.7246
0.4399	10.0	1290	0.7036	0.7434	0.7411
0.4347	11.0	1419	0.7512	0.7448	0.7290
0.4016	12.0	1548	0.7300	0.7393	0.7409
0.3724	13.0	1677	0.7419	0.7393	0.7401
0.3518	14.0	1806	0.7694	0.7352	0.7304
0.3401	15.0	1935	0.7626	0.7366	0.7344
0.3153	16.0	2064	0.8010	0.7407	0.7429
0.2976	17.0	2193	0.7879	0.7462	0.7439
0.2938	18.0	2322	0.7970	0.7476	0.7478
0.2732	19.0	2451	0.8209	0.7379	0.7390
0.2807	20.0	2580	0.8163	0.7490	0.7498

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.0
• Tokenizers 0.21.0