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: 1.6603
• Macro f1: 0.4329
• Weighted f1: 0.7053
• Accuracy: 0.7154
• Balanced accuracy: 0.4114

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: 3e-05
• train_batch_size: 32
• eval_batch_size: 32
• 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
1.3633	1.0	125	1.1325	0.3442	0.6470	0.6872	0.3862
1.0162	2.0	250	0.9858	0.3062	0.6889	0.7131	0.3135
0.868	3.0	375	0.9587	0.4091	0.7071	0.7207	0.3993
0.75	4.0	500	0.9983	0.4105	0.7080	0.7192	0.4039
0.6317	5.0	625	1.0197	0.4095	0.6941	0.6994	0.4093
0.5253	6.0	750	1.0760	0.4303	0.7073	0.7123	0.4223
0.4615	7.0	875	1.1371	0.4328	0.7040	0.7169	0.4096
0.3984	8.0	1000	1.1649	0.4516	0.6997	0.7002	0.4678
0.3332	9.0	1125	1.2009	0.4364	0.6994	0.7040	0.4243
0.2996	10.0	1250	1.2760	0.4336	0.7095	0.7192	0.4162
0.255	11.0	1375	1.3266	0.4353	0.6914	0.6918	0.4402
0.2318	12.0	1500	1.3591	0.4322	0.7011	0.7116	0.4101
0.2163	13.0	1625	1.4554	0.4226	0.7080	0.7237	0.4029
0.1837	14.0	1750	1.4363	0.4385	0.6938	0.6963	0.4250
0.1735	15.0	1875	1.5356	0.4363	0.7118	0.7230	0.4098
0.1526	16.0	2000	1.5731	0.4370	0.7073	0.7169	0.4181
0.1288	17.0	2125	1.6258	0.4406	0.7123	0.7245	0.4151
0.1321	18.0	2250	1.6590	0.4364	0.7081	0.7184	0.4148
0.114	19.0	2375	1.6598	0.4324	0.7074	0.7192	0.4081
0.1063	20.0	2500	1.6603	0.4329	0.7053	0.7154	0.4114

Framework versions

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