Coding Master Dataset: 766,987 Programming Instructions in ShareGPT Format

766,987 records form a large-scale instruction-tuning dataset for code generation, assembled from multiple open sources and deduplicated. The dataset was created by Voidreaper2026 and is formatted as JSONL in the ShareGPT conversational style. Its last recorded update was on May 6, 2026.

Use Cases

• Fine-tune code generation models based on conversational instruction data.
• Train models for programming feedback and correction based on the CodeFeedback-Filtered-Instruction subset.
• Develop secure coding assistants using examples from the secure_programming_dpo source.
• Benchmark instruction-following capabilities of language models on programming tasks.

Strengths

• 766,987 records provide a substantial volume of training examples.
• Assembled from multiple sources including CodeX-2M-Thinking (430,542 records) and python-code-dataset-500k (559,515 records), suggesting diversity.
• Deduplication process likely reduces redundancy.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the final merged dataset is provided, but the exact composition and overlap of source datasets is unclear.
• Freshness should be verified as the last update timestamp is from 2026.

Provenance

Source

Multiple open sources including CodeX-2M-Thinking, python-code-dataset-500k, StackPulse, CodeFeedback-Filtered-Instruction, and secure_programming_dpo.

Collection Method

Assembled from open sources and deduplicated.

Freshness

Last updated 2026-05-06 18:25:37