Upload Stack v2 sparse Python classes 36k snapshot

.gitattributes

@@ -58,3 +58,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 # Video files - compressed
 *.mp4 filter=lfs diff=lfs merge=lfs -text
 *.webm filter=lfs diff=lfs merge=lfs -text
+all.jsonl filter=lfs diff=lfs merge=lfs -text
+test.jsonl filter=lfs diff=lfs merge=lfs -text
+train.jsonl filter=lfs diff=lfs merge=lfs -text
+val.jsonl filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,51 @@
+---
+license: other
+task_categories:
+- text-generation
+language:
+- code
+- en
+pretty_name: Stack v2 Sparse Python Classes 36k
+size_categories:
+- 10K<n<100K
+---
+
+# Stack v2 Sparse Python Classes 36k
+
+This is a 36,000-sample snapshot for Diffusion + Autoregressive hybrid code generation experiments.
+
+## Source
+
+The data is extracted from `bigcode/the-stack-v2-dedup`, Python subset. The extraction uses Stack v2 metadata as source of truth, groups candidates by `repo_name + revision_id`, fetches files with git partial fetch + sparse checkout, then applies AST-level class filters.
+
+## Splits
+
+- `train.jsonl`: 26,000
+- `val.jsonl`: 5,000
+- `test.jsonl`: 5,000
+- `all.jsonl`: 36,000
+
+Note: the requested split `35k/5k/5k` sums to 45k, not 36k. This 36k snapshot keeps validation and test at 5k each and assigns the remaining 26k to train.
+
+## Record Format
+
+Each JSONL row is one Python class sample. Important fields include:
+
+- `prompt`: natural-language class implementation prompt
+- `skeleton`: class/method signatures and docstrings with `<|body_i|>` slots
+- `bodies`: list of method bodies without docstrings
+- `bodies_text`: body slots wrapped by `<|body_start_i|>` and `<|end_body_i|>`
+- `full_text`: skeleton plus body slots
+- `solution`: reconstructed class code
+- `source_repo`, `source_path`, `revision_id`, `blob_id`, `detected_licenses`: source metadata
+
+## Filters
+
+- 2 to 6 methods per class
+- every method has a non-empty docstring
+- every method body has 3 to 30 non-empty lines
+- reconstructed class parses as Python AST
+- tests/docs/examples/vendor/generated files are excluded by metadata/path filters
+- simple ClassEval/HumanEval contamination filters are applied
+
+Strict pyflakes is not used as a hard filter because isolated extracted classes often depend on module-level imports, constants, parent classes, or helper functions.

all.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06fc3f22d8fea4edfa0b2230f24110c75d1098184f7f603dbb52a9ffefcf72cc
+size 444140865

metadata.json

@@ -0,0 +1,119 @@
+{
+  "name": "stack_v2_sparse_classes_36k",
+  "source_dataset_dir": "data/stack_v2_sparse_classes_30k",
+  "source_dataset": "bigcode/the-stack-v2-dedup",
+  "source_data_files": "data/Python/*.parquet",
+  "snapshot_size": 36000,
+  "snapshot_seed": 42,
+  "bad_json_lines_skipped": 0,
+  "requested_split_note": "User requested a 36k dataset and also 35k/5k/5k splits, which sums to 45k. This 36k snapshot uses 26k/5k/5k to keep validation and test at 5k each.",
+  "splits": {
+    "train": 26000,
+    "val": 5000,
+    "test": 5000
+  },
+  "format": "class-level Python code generation with prompt, skeleton, method bodies, and reconstructed solution",
+  "filters": {
+    "min_methods": 2,
+    "max_methods": 6,
+    "require_method_docstrings": true,
+    "min_body_lines": 3,
+    "max_body_lines": 30,
+    "max_total_tokens": 1500,
+    "min_length_bytes": 500,
+    "max_length_bytes": 50000,
+    "skip_tests_docs_examples_init": true,
+    "pyflakes_hard_filter": false
+  },
+  "method_count_distribution": {
+    "2": 21570,
+    "3": 8008,
+    "4": 3586,
+    "5": 1856,
+    "6": 980
+  },
+  "top_repos": [
+    [
+      "cohesity/management-sdk-python",
+      456
+    ],
+    [
+      "microsoftgraph/msgraph-sdk-python",
+      401
+    ],
+    [
+      "home-assistant/core",
+      291
+    ],
+    [
+      "jansel/pytorch-jit-paritybench",
+      124
+    ],
+    [
+      "sungminoh/algorithms",
+      124
+    ],
+    [
+      "953250587/leetcode-python",
+      121
+    ],
+    [
+      "Shiv2157k/leet_code",
+      118
+    ],
+    [
+      "Jimmy-INL/google-research",
+      101
+    ],
+    [
+      "metoppv/improver",
+      91
+    ],
+    [
+      "bssrdf/pyleet",
+      90
+    ],
+    [
+      "dealflowteam/Idfy",
+      78
+    ],
+    [
+      "vincent-lg/tsunami",
+      76
+    ],
+    [
+      "oxhead/CodingYourWay",
+      75
+    ],
+    [
+      "vsdrun/lc_public",
+      75
+    ],
+    [
+      "JavaRod/SP_Python220B_2019",
+      75
+    ],
+    [
+      "mindspore-ai/models",
+      72
+    ],
+    [
+      "log2timeline/plaso",
+      68
+    ],
+    [
+      "inventree/InvenTree",
+      67
+    ],
+    [
+      "yiming1012/MyLeetCode",
+      64
+    ],
+    [
+      "luque/better-ways-of-thinking-about-software",
+      63
+    ]
+  ],
+  "avg_body_lines": 9.144970414201184,
+  "task": "Given class/method signatures and docstrings, generate class skeleton and method bodies for Diffusion + AR hybrid code generation experiments."
+}

test.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45bb3d8da6ea0f12a501f16316d4edfa56f3caaba3656c0e3b3a6844139806c1
+size 61490756

train.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfe4cc8c0d8451b01dbfdb0591e4bfad53f55e5b19f9784d0276a4847a09938d
+size 320962541

val.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bff1b6b656ddd819af37d18d505169ee8432ab0aed08156963202af330156a6f
+size 61687568