codebert-java-inconsistency

This model is a fine-tuned version of microsoft/codebert-base on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3543
• Accuracy: 0.9167
• F1: 0.9183
• Precision: 0.9235
• Recall: 0.9167

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: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
1.4625	3.1290	50	0.8954	0.7531	0.7554	0.7765	0.7531
0.5834	6.2581	100	0.5559	0.8189	0.8241	0.8483	0.8189
0.2858	9.3871	150	0.4046	0.8930	0.8945	0.8995	0.8930
0.1624	12.5161	200	0.4461	0.8642	0.8661	0.8750	0.8642
0.1084	15.6452	250	0.4012	0.9012	0.9038	0.9123	0.9012
0.074	18.7742	300	0.4689	0.8765	0.8817	0.8972	0.8765
0.0574	21.9032	350	0.4885	0.8807	0.8845	0.8970	0.8807
0.0452	25.0	400	0.4900	0.8848	0.8888	0.9011	0.8848
0.0396	28.1290	450	0.4896	0.8765	0.8805	0.8934	0.8765

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1