Add dataset card

README.md

@@ -0,0 +1,47 @@
+---
+license: mit
+pretty_name: Code2LoRA-Evo (evolution track, GRU)
+tags: [code, assertion-completion, repository-level]
+configs:
+- config_name: commits
+  data_files:
+  - {split: train, path: commits/train.parquet}
+  - {split: cr_val, path: commits/cr_val.parquet}
+  - {split: cr_test, path: commits/cr_test.parquet}
+- config_name: qna
+  data_files:
+  - {split: train, path: qna/train.parquet}
+  - {split: cr_val, path: qna/cr_val.parquet}
+  - {split: cr_test, path: qna/cr_test.parquet}
+  - {split: ir_val, path: qna/ir_val.parquet}
+  - {split: ir_test, path: qna/ir_test.parquet}
+---
+
+# Code2LoRA-Evo — evolution track (GRU)
+
+Everything the **Code2LoRA-Evo** GRU hypernetwork consumes, post-cap and
+self-contained. The GRU walks per-commit `diff_embedding` /
+`repo_state_embedding` and is trained with the smart-capped
+(`<=8`/commit, `<=4`/file) training QnAs; evaluation uses the paper's
+eval-time cap (first `<=8` QnAs per (repo, commit)).
+
+| config | split | rows | role |
+|---|---|---:|---|
+| commits | train | 58,617 | per-commit diff + repo-state embeddings |
+| commits | cr_val / cr_test | 8,614 / 6,618 | held-out repo commit walks |
+| qna | train | 215,129 | smart-capped training loss (`Code2LoRA-Evo`) |
+| qna | cr_val | 58,944 | capped eval (`<=8`/commit) |
+| qna | cr_test | 44,732 | capped eval |
+| qna | ir_val | 38,783 | capped eval |
+| qna | ir_test | 42,061 | capped eval |
+
+Join `commits` and `qna` on `(repo_id, commit_sha)`.
+
+```python
+from datasets import load_dataset
+commits = load_dataset("code2lora/code2lora-evo", "commits")
+qna     = load_dataset("code2lora/code2lora-evo", "qna")
+```
+
+Companion datasets: `code2lora/code2lora-static`,
+`code2lora/code2lora-static-anchor`, `code2lora/repopeftbench-ood`.