datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
csharp-dotnet-cpt-v0
csharp-dotnet-cpt-v0
A reproducible C# / .NET continued-pretraining (CPT) corpus for Qwen2.5-Coder-1.5B.
HF: https://huggingface.co/datasets/Nottybro/csharp-dotnet-cpt-v0 (private)
Total unique tokens: 695,288,168 (Qwen2.5-Coder-1.5B tokenizer)
Files: 681,426 | Repositories: 68,869
Format: Zstandard-compressed Parquet, schema below.
See DATASET_CARD.md for sources, license policy, filtering, limitations.
See reports/summary.md for full statistics.
dotnet-runtime
.NET Runtime Fine-Tuning Data and Index
This directory contains data for fine-tuning models and building RAGs for the dotnet/runtime repository.
Overview
data/: Contains all datasets and indexes.
raw/sample/: Sample PRs and diffs collected from GitHub.
raw_data.tar: Archive of collected PRs and diffs from GitHub.
samples/: Json files with processed samples suitable for dataset generation.
processed/: Parquet files for fine-tuning (e.g., train.parquet, test.parquet).… See the full description on the dataset page: https://huggingface.co/datasets/kotlarmilos/dotnet-runtime.