github_issues-dataset-distilbert-base-uncased

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

• Loss: 0.1495
• Accuracy: 0.9580
• F1: 0.6067
• Precision: 0.7297
• Recall: 0.5192

Model description

distilbert-base-uncased

Intended uses & limitations

Multi Label Classification on GitHub repository issues.

Training and evaluation data

GitHub issues dataset taken from GitHub issues.

Split the dataset into 80-20 train-test splits. Filtered out the pull requests and issues with no labels.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.3962	1.0	114	0.2513	0.9208	0.34	0.3542	0.3269
0.2008	2.0	228	0.1847	0.9436	0.4198	0.5862	0.3269
0.1633	3.0	342	0.1608	0.9544	0.5581	0.7059	0.4615
0.1468	4.0	456	0.1519	0.9580	0.6067	0.7297	0.5192
0.1385	5.0	570	0.1495	0.9580	0.6067	0.7297	0.5192

Framework versions

• Transformers 4.49.0
• Pytorch 2.6.0+cu124
• Datasets 3.4.1
• Tokenizers 0.21.1