Langmap Thestack Python 100M: Python Code Finetuning Dataset

LangMap-TheStack-python-100M is a dataset of 100 million tokens of sanitized Python source code, intended for code finetuning. It was created by MultilingualUnigramLM and streamed from the bigcode/the-stack repository. The dataset was last updated on April 13, 2026.

Use Cases

• Fine-tuning language models for Python code generation based on the sanitized source code.
• Training models for code completion tasks using the tokenized Python corpus.
• Evaluating the performance of code-specific tokenizers like allenai/OLMo-3-1025-7B.
• Studying patterns in large-scale Python code repositories sourced from The Stack.

Strengths

• Contains 100 million tokens of Python code, meeting a specific target size.
• Source code is sanitized, which likely reduces noise and formatting issues.
• Uses a specific tokenizer (allenai/OLMo-3-1025-7B), providing consistency for model training.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file size are unknown, which may limit suitability assessment.
• The dataset's license is unknown, which could restrict usage.

Provenance

Source

bigcode/the-stack repository on Hugging Face.

Collection Method

Streamed and processed for finetuning.

Freshness

Last updated 2026-04-13 19:17:09; freshness should be verified.