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: 2.4011
• Macro f1: 0.3527
• Weighted f1: 0.6956
• Accuracy: 0.7177
• Balanced accuracy: 0.3299

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: 5e-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
1.2992	1.0	250	1.1977	0.1984	0.6212	0.6979	0.2104
1.1076	2.0	500	1.0809	0.2865	0.6479	0.6986	0.2924
0.912	3.0	750	1.1359	0.2677	0.6718	0.6804	0.2882
0.7969	4.0	1000	1.1522	0.2643	0.6840	0.7047	0.2692
0.6313	5.0	1250	1.2438	0.3176	0.6856	0.6986	0.3149
0.542	6.0	1500	1.3582	0.3212	0.6736	0.6872	0.3173
0.4401	7.0	1750	1.4300	0.3472	0.6921	0.7024	0.3305
0.382	8.0	2000	1.5530	0.3669	0.6965	0.7146	0.3480
0.309	9.0	2250	1.7972	0.3390	0.6777	0.6986	0.3174
0.2762	10.0	2500	1.7713	0.3745	0.6923	0.7161	0.3396
0.242	11.0	2750	1.9214	0.3672	0.6982	0.7215	0.3373
0.2112	12.0	3000	1.9624	0.3543	0.6917	0.7093	0.3310
0.179	13.0	3250	2.0087	0.3658	0.6922	0.7078	0.3431
0.1563	14.0	3500	2.1266	0.3554	0.7016	0.7237	0.3331
0.1531	15.0	3750	2.2341	0.3479	0.6951	0.7123	0.3284
0.115	16.0	4000	2.2671	0.3565	0.6970	0.7207	0.3308
0.115	17.0	4250	2.3446	0.3547	0.6988	0.7199	0.3342
0.0931	18.0	4500	2.3784	0.3570	0.6977	0.7169	0.3333
0.0886	19.0	4750	2.3871	0.3557	0.6970	0.7169	0.3325
0.0747	20.0	5000	2.4011	0.3527	0.6956	0.7177	0.3299

Framework versions

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