bert-base-uncased-finetuned-sdg

This model is a fine-tuned version of bert-base-uncased on the OSDG dataset. It achieves the following results on the evaluation set:

• Loss: 0.3094
• Acc: 0.9195

Model description

Classifies text to the first 16 SDGs!

Intended uses & limitations

Assess policy documents, classify text to SDGs, etc.

Training and evaluation data

OSDG data. Updated version from October.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 64
• eval_batch_size: 64
• seed: 42
• distributed_type: multi-GPU
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Acc
0.3768	1.0	269	0.3758	0.8933
0.2261	2.0	538	0.3088	0.9095
0.1038	3.0	807	0.3094	0.9195

Framework versions

• Transformers 4.23.1
• Pytorch 1.12.0a0+8a1a93a
• Datasets 2.5.2
• Tokenizers 0.13.1