by MoreThought
DeepSWEGym2 Edu is MoreThought's filtered and deduplicated merge of several software-engineering datasets, with 53,649 examples of long, complex code problems. It is meant for training coding agents and is MIT licensed.
DatasetFree
Step 1: Open the dataset page on Hugging Face and review the source datasets listed in the card.
Step 2: Load the dataset with the Hugging Face datasets library.
Step 3: Plan for very large examples, since some reach 28 MB, and use specialized training scripts.
Step 4: Avoid mixing it with other variants or versions of the dataset to prevent overlapping examples.
Step 5: Train or fine-tune your coding model on the data.
MIT, according to the dataset card.
It has 53,649 examples and is about 14.93 GB uncompressed, with an average row size of 291.74 kb.
The card says not to, if you want to avoid overlapping examples.
Yes. Examples can reach 28 MB, which almost all language models cannot handle directly, so the card recommends specialized scripts.
GSM8KGSM8K is OpenAI's dataset of about 8.5K grade-school math word problems with step-by-step natural-language solutions. It is for researchers testing and training language models on multi-step arithmetic reasoning.