Skip to content

Dataset · Modotte

CodeX-2M-Thinking Synthetic SFT Dataset

Modotte/CodeX-2M-Thinking

Synthetic English texts for instruction tuning and fine-tuning response generation models.

What's inside

The dataset is fully synthetic: compiled from high-quality public sources and augmented with data generated by closed and open models. This is part of the CodeX line from Modotte, focused on instruction-based tuning and fine-tuning. The language is English, with a volume of 1 million to 10 million examples.

What you can do with it

  • Fine-tuning an instruct model on synthetic SFT data
  • Configuring LoRA for response generation
  • Evaluating the quality of the synthetic dataset for fine-tuning
Tasks
Text generation, Question answering
Languages
English
Size
1M–10M rows
Downloads this month
895
Likes
131

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("Modotte/CodeX-2M-Thinking")