README.md

metadata

tags:
  - generated_from_trainer
metrics:
  - accuracy
  - f1
model-index:
  - name: CommitPredictor
    results: []

CommitPredictor

This model is a fine-tuned version of microsoft/codebert-base-mlm on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.5888
• Accuracy: 0.8783
• F1: 0.8783
• Bleu4: 0.8598

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: 42
• eval_batch_size: 42
• seed: 42
• gradient_accumulation_steps: 3
• total_train_batch_size: 126
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Bleu4
No log	1.0	236	0.8706	0.8253	0.8253	0.7764
No log	2.0	472	0.7296	0.8503	0.8503	0.8287
1.0825	3.0	708	0.6826	0.8594	0.8594	0.8123
1.0825	4.0	944	0.6655	0.8645	0.8645	0.8480
0.755	5.0	1180	0.6317	0.8696	0.8696	0.9028
0.755	6.0	1416	0.6333	0.8699	0.8699	0.8870
0.6948	7.0	1652	0.6147	0.8738	0.8738	0.9187
0.6948	8.0	1888	0.6110	0.8738	0.8738	0.8080
0.6633	9.0	2124	0.5987	0.8770	0.8770	0.8903
0.6633	10.0	2360	0.5888	0.8783	0.8783	0.8598

Framework versions

• Transformers 4.25.1
• Pytorch 1.13.0+cu117
• Datasets 2.7.1
• Tokenizers 0.13.2