curelink-biomed-nli-v3

This model is a fine-tuned version of CureLink/curelink-biomed-nli-v3 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.4855
• Accuracy: 0.865
• F1 Macro: 0.8650

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: 2e-05
• train_batch_size: 8
• eval_batch_size: 16
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 200
• num_epochs: 4

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1 Macro
0.9051	0.12	450	0.4044	0.848	0.8470
0.9096	0.24	900	0.3758	0.858	0.8580
0.8731	0.36	1350	0.3897	0.847	0.8463
0.9115	0.48	1800	0.4021	0.842	0.8416
0.8851	0.6	2250	0.3811	0.8515	0.8512
0.8686	0.72	2700	0.3943	0.8495	0.8497
0.8941	0.84	3150	0.3857	0.851	0.8502
0.8890	0.96	3600	0.4030	0.847	0.8459
0.7496	1.08	4050	0.4119	0.854	0.8541
0.7359	1.2	4500	0.3885	0.854	0.8533
0.6676	1.32	4950	0.4190	0.8535	0.8527
0.7652	1.44	5400	0.3823	0.8575	0.8570
0.7476	1.56	5850	0.3856	0.8555	0.8556
0.7203	1.6800	6300	0.4120	0.854	0.8538
0.7375	1.8	6750	0.4043	0.853	0.8530
0.6821	1.92	7200	0.3879	0.857	0.8564
0.6303	2.04	7650	0.4565	0.862	0.8616
0.5527	2.16	8100	0.4573	0.857	0.8570
0.5964	2.2800	8550	0.4442	0.8615	0.8611
0.5569	2.4	9000	0.4403	0.859	0.8594
0.5861	2.52	9450	0.4435	0.8595	0.8600
0.5866	2.64	9900	0.4500	0.8635	0.8635
0.5657	2.76	10350	0.4446	0.8605	0.8608
0.6038	2.88	10800	0.4253	0.8625	0.8627
0.6190	3.0	11250	0.4115	0.868	0.8684
0.5386	3.12	11700	0.4718	0.8645	0.8649
0.5231	3.24	12150	0.4877	0.864	0.8644
0.4741	3.36	12600	0.4825	0.8655	0.8655
0.5221	3.48	13050	0.4838	0.8685	0.8687
0.5286	3.6	13500	0.4840	0.8625	0.8622
0.5310	3.7200	13950	0.4832	0.867	0.8672
0.5495	3.84	14400	0.4878	0.8635	0.8635
0.4940	3.96	14850	0.4860	0.865	0.8650
0.5326	4.0	15000	0.4855	0.865	0.8650

Framework versions

• Transformers 5.5.0
• Pytorch 2.11.0
• Datasets 4.8.4
• Tokenizers 0.22.2