Exclude some additional schemas

data/test.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f2e81c3efee377cc879ea4c839e18e856ce36df1ea9638aad4ae2907438ee44a
-size 3540592
+oid sha256:e29f0f20dba83e0a15297f0411ec1d8644b174bb48fcdea0ce1d83d0abb92c40
+size 1118078

data/train.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6bbdc698f998f9665f296041dcab6fa5d44177a0fb532d8be3193d849a97a2f6
-size 10943729
+oid sha256:07aa22ac52077d44ab69d39efc391414707f21bb3ea49ded7a0bd477efebd9f0
+size 14192760

data/validation.jsonl.gz

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f423bd6a83aecc03e4ff761dbf1f14a4cd2c020c5a9cc53e90f9950f865f4b31
-size 2745523
+oid sha256:80a44f0d51e8215789f6288f56e9a6da45d57e794f772084797a28e6445ce90c
+size 1807245

train_split.py

@@ -22,9 +22,7 @@ def files_list(licenses):
     files = [
         f
         for f in data_path.rglob("*.json")
-        if f.is_file()
-        and "node_modules" not in f.parts
-        and licenses["/".join(f.parts[1:3])] in PERMISSIVE_LICENSES
+        if f.is_file() and licenses["/".join(f.parts[1:3])] in PERMISSIVE_LICENSES
     ]
     return files
 
@@ -164,7 +162,7 @@ def main(similarity, split, seed, commits_file, licenses_file, languages_file):
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument("--similarity", default=None, type=float)
-    parser.add_argument("--seed", default=282, type=int)
+    parser.add_argument("--seed", default=38, type=int)
     parser.add_argument("--split", default=0.8, type=float)
     parser.add_argument("--commits_file", default="commits.json")
     parser.add_argument("--licenses_file", default="licenses.json")

validate_schemas.py

@@ -8,6 +8,16 @@ import jsonschema
 from tqdm.contrib.concurrent import process_map
 
 
+IGNORE_PATHS = [
+    "node_modules",
+    "site-packages",
+    "draft2019-09",
+    "draft2020-12",
+    "draft-next",
+    "vendor",
+]
+
+
 def process_file(schema_file):
     # Calculate the path of the new file
     new_schema_file = Path("valid_data", *schema_file.parts[1:])
@@ -16,11 +26,20 @@ def process_file(schema_file):
     if not schema_file.is_file() and not new_schema_file.is_file():
         return
 
+    # Skip files in ignored directories
+    for path in IGNORE_PATHS:
+        if path in schema_file.parts or path in new_schema_file.parts:
+            return
+
     try:
         schema = json5.load(open(schema_file))
     except ValueError:
         return
 
+    # Skip meta schemas
+    if schema.get("$id").startswith("https://json-schema.org/"):
+        continue
+
     vcls = jsonschema.validators.validator_for(schema)
     try:
         vcls.check_schema(schema)