README.md

metadata

base_model: allenai/scibert_scivocab_uncased
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: scibert_prefix_cont_ll_SEP
    results: []

scibert_prefix_cont_ll_SEP

This model is a fine-tuned version of allenai/scibert_scivocab_uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0769
• F1 Weighted: 0.9112
• F1 Samples: 0.9155
• F1 Macro: 0.8184
• F1 Micro: 0.9121
• Accuracy: 0.8863

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: 15

Training results

Training Loss	Epoch	Step	Validation Loss	F1 Weighted	F1 Samples	F1 Macro	F1 Micro	Accuracy
0.2213	0.3381	500	0.1392	0.8151	0.8223	0.6081	0.8355	0.8018
0.1377	0.6761	1000	0.1129	0.8523	0.8584	0.6889	0.8645	0.8342
0.1214	1.0142	1500	0.1103	0.8504	0.8552	0.6955	0.8613	0.8302
0.0921	1.3523	2000	0.0961	0.8656	0.8655	0.7111	0.8740	0.8390
0.0863	1.6903	2500	0.0900	0.8789	0.8810	0.7281	0.8847	0.8545
0.0825	2.0284	3000	0.0959	0.8764	0.8844	0.7323	0.8826	0.8532
0.0567	2.3665	3500	0.0856	0.8879	0.8951	0.7454	0.8922	0.8633
0.061	2.7045	4000	0.0952	0.8802	0.8827	0.7397	0.8856	0.8586
0.0532	3.0426	4500	0.0839	0.8979	0.9058	0.7639	0.9031	0.8775
0.0361	3.3807	5000	0.0831	0.9007	0.9113	0.7791	0.9045	0.8769
0.0369	3.7187	5500	0.0833	0.9018	0.9094	0.7880	0.9031	0.8775
0.0392	4.0568	6000	0.0826	0.9062	0.9108	0.8180	0.9081	0.8823
0.027	4.3949	6500	0.0769	0.9112	0.9155	0.8184	0.9121	0.8863
0.0251	4.7329	7000	0.0868	0.8996	0.9061	0.7693	0.9018	0.8714
0.0255	5.0710	7500	0.0867	0.9083	0.9147	0.8048	0.9115	0.8870
0.0212	5.4091	8000	0.0834	0.9100	0.9161	0.8209	0.9116	0.8850

Framework versions

• Transformers 4.41.2
• Pytorch 2.3.1+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1

Citation

If using this model in your work, please cite:

@misc {scibert_claim-classification,
    author       = { Bleuze, Clémentine },
    title        = { Fine-tuned SciBERT model for claim classification  },
    year         = 2024,
    url          = { https://huggingface.co/ClementineBleuze/scibert_claim-classification },
    doi          = { 10.57967/hf/4797 },
    publisher    = { Hugging Face }
}