README.md

metadata

license: mit
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: MiniLM-evidence-types
    results: []

MiniLM-evidence-types

This model is a fine-tuned version of microsoft/MiniLM-L12-H384-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8388
• Macro f1: 0.4307
• Weighted f1: 0.6983
• Accuracy: 0.7032
• Balanced accuracy: 0.4139

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: 16
• eval_batch_size: 16
• 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	Macro f1	Weighted f1	Accuracy	Balanced accuracy
1.3124	1.0	250	1.1166	0.2582	0.6393	0.6788	0.2758
0.9939	2.0	500	0.9671	0.3859	0.6988	0.7093	0.3799
0.8486	3.0	750	1.0263	0.3519	0.6632	0.6606	0.3642
0.7396	4.0	1000	1.0125	0.4195	0.7092	0.7192	0.4186
0.6425	5.0	1250	1.0983	0.3910	0.6746	0.6826	0.3925
0.5648	6.0	1500	1.0948	0.4184	0.7145	0.7222	0.4089
0.4858	7.0	1750	1.1658	0.4242	0.7058	0.7184	0.4279
0.4329	8.0	2000	1.3020	0.4178	0.6806	0.6849	0.4081
0.3799	9.0	2250	1.2622	0.4466	0.7004	0.7055	0.4419
0.326	10.0	2500	1.3822	0.4162	0.6971	0.7032	0.4048
0.2849	11.0	2750	1.4716	0.3933	0.6941	0.6971	0.3826
0.251	12.0	3000	1.5651	0.4259	0.6928	0.6956	0.4231
0.2205	13.0	3250	1.6920	0.4257	0.6942	0.7032	0.4112
0.205	14.0	3500	1.7016	0.4269	0.6899	0.6872	0.4260
0.1946	15.0	3750	1.7647	0.4312	0.6891	0.6910	0.4232
0.1661	16.0	4000	1.8255	0.4168	0.6886	0.6933	0.4003
0.1502	17.0	4250	1.8261	0.4190	0.6950	0.7040	0.3996
0.1625	18.0	4500	1.8163	0.4260	0.7001	0.7047	0.4079
0.1329	19.0	4750	1.8274	0.4368	0.7023	0.7055	0.4218
0.1248	20.0	5000	1.8388	0.4307	0.6983	0.7032	0.4139

Framework versions

• Transformers 4.19.2
• Pytorch 1.11.0+cu113
• Datasets 2.2.2
• Tokenizers 0.12.1