dnabert2_ft_BioS74_1kbpHG19_DHSs_H3K27AC

This model is a fine-tuned version of vivym/DNABERT-2-117M on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.4505
• F1 Score: 0.8221
• Precision: 0.7638
• Recall: 0.8900
• Accuracy: 0.7983
• Auc: 0.8676
• Prc: 0.8468

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Score	Precision	Recall	Accuracy	Auc	Prc
0.5891	0.1314	500	0.5389	0.7519	0.7628	0.7413	0.7439	0.8160	0.8026
0.5602	0.2629	1000	0.5736	0.7134	0.7884	0.6514	0.7260	0.8191	0.8066
0.5579	0.3943	1500	0.5285	0.7623	0.7524	0.7725	0.7478	0.8252	0.8155
0.5246	0.5258	2000	0.5114	0.7773	0.7420	0.8162	0.7552	0.8290	0.8167
0.5286	0.6572	2500	0.5336	0.7872	0.7000	0.8990	0.7455	0.8325	0.8193
0.5365	0.7886	3000	0.6523	0.7035	0.8083	0.6228	0.7252	0.8170	0.8151
0.5339	0.9201	3500	0.5058	0.7904	0.7253	0.8684	0.7589	0.8360	0.8243
0.5288	1.0515	4000	0.5916	0.7392	0.7945	0.6911	0.7447	0.8348	0.8223
0.5279	1.1830	4500	0.5182	0.7912	0.7099	0.8935	0.7531	0.8372	0.8266
0.5273	1.3144	5000	0.5405	0.7770	0.7710	0.7830	0.7647	0.8383	0.8288
0.5251	1.4458	5500	0.4926	0.7811	0.7653	0.7976	0.7660	0.8400	0.8282
0.5126	1.5773	6000	0.5148	0.7756	0.7736	0.7775	0.7644	0.8401	0.8288
0.5102	1.7087	6500	0.4890	0.7933	0.7344	0.8624	0.7647	0.8412	0.8302
0.509	1.8402	7000	0.5007	0.7833	0.7609	0.8071	0.7662	0.8418	0.8313
0.5116	1.9716	7500	0.4818	0.7815	0.7745	0.7885	0.7691	0.8466	0.8363
0.5065	2.1030	8000	0.5003	0.7920	0.6900	0.9292	0.7444	0.8476	0.8384
0.4948	2.2345	8500	0.4960	0.8037	0.7134	0.9201	0.7647	0.8481	0.8373
0.5005	2.3659	9000	0.4866	0.7677	0.8020	0.7363	0.7668	0.8522	0.8417
0.4773	2.4974	9500	0.4903	0.8040	0.7147	0.9186	0.7654	0.8533	0.8429
0.4816	2.6288	10000	0.4762	0.8013	0.7680	0.8378	0.7825	0.8577	0.8476
0.4907	2.7603	10500	0.5126	0.7936	0.7858	0.8016	0.7818	0.8608	0.8505
0.4847	2.8917	11000	0.4826	0.7666	0.8061	0.7308	0.7670	0.8574	0.8499
0.465	3.0231	11500	0.5577	0.8013	0.6940	0.9478	0.7539	0.8640	0.8548
0.461	3.1546	12000	0.4721	0.8111	0.7221	0.9252	0.7744	0.8665	0.8563
0.4422	3.2860	12500	0.4777	0.8133	0.7523	0.8850	0.7873	0.8675	0.8585
0.4524	3.4175	13000	0.4947	0.7972	0.7843	0.8106	0.7841	0.8650	0.8569
0.457	3.5489	13500	0.4511	0.8044	0.7801	0.8302	0.7886	0.8690	0.8591
0.455	3.6803	14000	0.4756	0.8085	0.7115	0.9362	0.7678	0.8649	0.8539
0.4611	3.8118	14500	0.4829	0.8097	0.7134	0.9362	0.7697	0.8694	0.8611
0.4635	3.9432	15000	0.4506	0.8139	0.7605	0.8754	0.7904	0.8688	0.8598
0.4475	4.0747	15500	0.4597	0.8120	0.7805	0.8463	0.7949	0.8739	0.8652
0.433	4.2061	16000	0.4842	0.8202	0.7438	0.9141	0.7902	0.8590	0.8405
0.446	4.3375	16500	0.4672	0.8198	0.7409	0.9176	0.7889	0.8747	0.8656
0.4242	4.4690	17000	0.4723	0.8229	0.7548	0.9046	0.7962	0.8723	0.8672
0.4432	4.6004	17500	0.4880	0.8143	0.7193	0.9382	0.7760	0.8723	0.8657
0.4446	4.7319	18000	0.4821	0.8186	0.7372	0.9201	0.7865	0.8756	0.8683
0.4353	4.8633	18500	0.4721	0.8191	0.7775	0.8654	0.7999	0.8754	0.8667
0.4238	4.9947	19000	0.4882	0.8017	0.8074	0.7961	0.7938	0.8744	0.8663
0.4207	5.1262	19500	0.4505	0.8221	0.7638	0.8900	0.7983	0.8676	0.8468

Framework versions

• Transformers 4.46.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 2.18.0
• Tokenizers 0.20.0