LongSpark
Collection
4 items • Updated • 2
Datasets:
id stringlengths 17 17 | language stringclasses 5
values | prompt stringlengths 189k 334k | input_ids listlengths 65.5k 65.5k | source dict |
|---|---|---|---|---|
codespan_64K_0001 | C | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,3284,6126,17(...TRUNCATED) | {"repository":"mackron/dr_libs","path":"dr_flac.h","revision":"dfe8377631000664666519fdb83da193fd803(...TRUNCATED) |
codespan_64K_0002 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,322,24253,49(...TRUNCATED) | {"repository":"ocornut/imgui","path":"imgui_demo.cpp","revision":"e8602501bb900fc7c965e65b88fef30b2c(...TRUNCATED) |
codespan_64K_0003 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,322,24253,49(...TRUNCATED) | {"repository":"ocornut/imgui","path":"imgui_widgets.cpp","revision":"e8602501bb900fc7c965e65b88fef30(...TRUNCATED) |
codespan_64K_0004 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,322,24253,49(...TRUNCATED) | {"repository":"ocornut/imgui","path":"imgui.cpp","revision":"e8602501bb900fc7c965e65b88fef30b2c80aff(...TRUNCATED) |
codespan_64K_0005 | C | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,1057,357,65,(...TRUNCATED) | {"repository":"nothings/stb","path":"stb_image_resize2.h","revision":"2c980bb59875b0d32144a71867fbde(...TRUNCATED) |
codespan_64K_0006 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,322,24253,49(...TRUNCATED) | {"repository":"ocornut/imgui","path":"imgui_draw.cpp","revision":"e8602501bb900fc7c965e65b88fef30b2c(...TRUNCATED) |
codespan_64K_0007 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,78712,4421,3(...TRUNCATED) | {"repository":"llvm/llvm-project","path":"clang/lib/Sema/SemaDecl.cpp","revision":"llvmorg-20.1.8","(...TRUNCATED) |
codespan_64K_0008 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,78712,4421,3(...TRUNCATED) | {"repository":"llvm/llvm-project","path":"clang/lib/Sema/SemaExpr.cpp","revision":"llvmorg-20.1.8","(...TRUNCATED) |
codespan_64K_0009 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,1057,356,102(...TRUNCATED) | {"repository":"gcc-mirror/gcc","path":"gcc/cp/parser.cc","revision":"e7ec537b048c89310c80b30593bf0ff(...TRUNCATED) |
codespan_64K_0010 | C++ | "<|im_start|>user\nContinue the following document with a coherent next passage. Output only the con(...TRUNCATED) | [151644,872,198,23526,279,2701,2197,448,264,55787,1790,21085,13,9258,1172,279,41171,382,1057,13760,5(...TRUNCATED) | {"repository":"gcc-mirror/gcc","path":"gcc/cp/pt.cc","revision":"e7ec537b048c89310c80b30593bf0ff29e2(...TRUNCATED) |
End of preview. Expand in Data Studio
CodeSpan
64K source-code continuation for long-context decoding.
CodeSpan contains 32 examples from 17 open-source projects, including LLVM, GCC, Linux, and PostgreSQL. Each example is a contiguous prefix of a distinct source file, wrapped as a continuation prompt: 65,536 Qwen3 tokens, including the chat template with thinking disabled. Files are neither concatenated nor repeated. There are at most four files per project; 28 of the 32 files are C/C++.
| Configuration | Contents |
|---|---|
default |
id, language, formatted prompt, input_ids, and source with repository, revision, path, URL, license, and modification notice. |
longspark |
The same 32 examples in the same order, as source_id and input_ids, matching the LongSpark evaluation inputs. |
from datasets import load_dataset
data = load_dataset("Hehy/CodeSpan", split="test")
input_ids = data[0]["input_ids"]
Use the supplied token IDs directly; the chat template is already included. This is a decoding workload with no reference continuations. The release manifest records file hashes and tokenizer identity.
Source excerpts retain their upstream licenses. Full license texts, copyright notices, and the per-sample inventory are in THIRD_PARTY_NOTICES.txt; retain them when redistributing the data.
- Downloads last month
- 29