CodeGeneration-CodeT5-base

This model was trained from the pretrained CodeT5-base model on ThoughtFocusAI/Python-CodeGeneration dataset. It achieves the following results on the evaluation set:

• ngram match: 0.014948515566713798
• weighted ngram match: 0.017803898905793227
• syntax_match: 0.21961325966850828
• dataflow_match: 0.36255924170616116
• bleu = 1.87
• codebleu = 15.3731
• exact match = 0.0

Model description

This model was trained from the pretrained CodeT5-base model Salesforce and has been fine tuned for Python Code Genertion.

Intended uses & limitations

Python Code Generation. Limited to simple queries.

Training and evaluation data

ThoughtFocusAI/Python-CodeGeneration

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• optimizer: None
• training_precision: float32

Training results

Framework versions

• Transformers 4.29.2
• TensorFlow 2.12.0
• Datasets 2.12.0
• Tokenizers 0.13.3