File size: 955 Bytes
1378d9b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 |
Codegen is a model for conversational program synthesis, where each problem is interactively solved in multiple steps, each consisting of a natural language specification from the user and a synthesized subprogram from the system.
It was was sequentially trained on three datasets:
- The Pile
- A 341GB subset of Google’s [BigQuery dataset](https://cloud.google.com/blog/topics/public-datasets/github-on-bigquery-analyze-all-the-open-source-code) of code files from multiple programming languages, keeping only 6: C, C++, Go, Java, JavaScript, and Python
- 217GB of Python data from Github repositories
The second and third datasets used the following preprocessing:
- Exact match deduplication
- Filtering:
- Exact match deduplication
- Average line length < 100 tokens
- Maximum line length < 1000 MB
- >90% of the characters being decimal or hexadecimal digits
**Remark**:
The reported data sizes are after preprocessing. |