Add balanced repository-level split near 80 10 10

README.md

@@ -8,6 +8,14 @@ configs:
     path: repo_split/validation-*
   - split: test
     path: repo_split/test-*
+- config_name: repo_split_balanced
+  data_files:
+  - split: train
+    path: repo_split_balanced/train-*
+  - split: validation
+    path: repo_split_balanced/validation-*
+  - split: test
+    path: repo_split_balanced/test-*
 - config_name: random_shuffle
   data_files:
   - split: train
@@ -30,6 +38,7 @@ This dataset keeps the column format of `hreyulog/arkts-code-docstring` and cont
 It provides two configurations:
 
 - `repo_split`: train/validation/test are split by repository, so a repository appears in only one split.
+- `repo_split_balanced`: repositories are kept isolated while repository assignments are weighted by example count to approximate an 80%/10%/10% example split.
 - `random_shuffle`: all examples are shuffled with seed `20260714` and then split 80%/10%/10%.
 
 Columns are: `nwo`, `sha`, `path`, `language`, `identifier`, `docstring`, `function`, `ast_function`, `obf_function`, `url`, `function_sha`, and `source`.

manifest.json

@@ -20,6 +20,11 @@
       "validation": 9207,
       "train": 45200
     },
+    "repo_split_balanced": {
+      "train": 47493,
+      "validation": 5936,
+      "test": 5936
+    },
     "random_shuffle": {
       "test": 5936,
       "validation": 5937,

repo_split_balanced/test-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bec780f194de135c9ee5387c578b64a4514fa91ec205b76bf00562e0516519b
+size 4947628

repo_split_balanced/train-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e342b98189f6dd41a61695e0a476718ff80e81c65410456beaaf224d01212e29
+size 34899784

repo_split_balanced/validation-00000-of-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c516da378c0f8f4a8ff77e319e680826e585cf88b177080db511d666760ea596
+size 4958662