verifiers 0.4.2.dev28__tar.gz → 0.4.2.dev29__tar.gz

{verifiers-0.4.2.dev28 → verifiers-0.4.2.dev29}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: verifiers
-Version: 0.4.2.dev28
+Version: 0.4.2.dev29
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers

{verifiers-0.4.2.dev28 → verifiers-0.4.2.dev29}/tests/unit/test_taskset.py

@@ -102,6 +102,16 @@ def test_exclude_drops_tasks_that_any_list_names() -> None:
     assert idxs(finite().exclude(idx=":7", names=["t8"])) == [7, 9]
 
 
+def test_match_lists_extend_from_files(tmp_path) -> None:
+    (tmp_path / "idx.txt").write_text("0:2\n\n5\n")
+    (tmp_path / "names.txt").write_text(" t8 \nt9\n")
+    files = {"idx": tmp_path / "idx.txt", "names": tmp_path / "names.txt"}
+    assert idxs(finite().include(idx=[3], files=files)) == [0, 1, 3, 5, 8, 9]
+    config = vf.SelectConfig(exclude={"files": {"names": tmp_path / "names.txt"}})
+    assert config.exclude.names == ["t8", "t9"] and config.exclude.files.names is None
+    assert idxs(finite().select(config)) == list(range(8))
+
+
 def test_unmatched_entries_warn(caplog) -> None:
     with caplog.at_level(logging.WARNING):
         assert idxs(finite().include(names=["t1", "typo"], idx=[42])) == [1]

{verifiers-0.4.2.dev28 → verifiers-0.4.2.dev29}/verifiers/v1/configs/select.py

@@ -3,13 +3,24 @@
 
 import re
 import sys
+from pathlib import Path
 
-from pydantic import AliasChoices, Field, field_validator
+from pydantic import AliasChoices, Field, field_validator, model_validator
 from pydantic_config import BaseConfig
 
 IDX_RANGE = re.compile(r"(\d*):(\d*)(?::(\d*))?")
 
 
+class TaskMatchFilesConfig(BaseConfig):
+    """Newline-delimited text files with more entries for each `TaskMatchConfig`
+    list, one entry per line. Blank lines are skipped."""
+
+    idx: Path | None = None
+    ids: Path | None = None
+    keys: Path | None = None
+    names: Path | None = None
+
+
 class TaskMatchConfig(BaseConfig):
     """Tasks named by position or identity; a task matches when any list names it."""
 
@@ -24,6 +35,10 @@ class TaskMatchConfig(BaseConfig):
     """`Task.key` values, as recorded on each trace (`task.key`)."""
     names: list[str] = Field(default_factory=list)
     """`TaskData.name` values."""
+    files: TaskMatchFilesConfig = TaskMatchFilesConfig()
+    """Files whose lines extend the lists above, for selections too long to write
+    inline (`files.ids = "ids.txt"`). Validation reads them into the lists and
+    resets `files`."""
 
     @field_validator("idx", mode="before")
     @classmethod
@@ -36,6 +51,19 @@ class TaskMatchConfig(BaseConfig):
             items.extend(_parse_idx_item(part) for part in parts)
         return items
 
+    @model_validator(mode="after")
+    def _read_files(self):
+        for field, path in self.files:
+            if path is None:
+                continue
+            lines = [line.strip() for line in path.read_text().splitlines()]
+            entries = [line for line in lines if line]
+            if field == "idx":
+                entries = [_parse_idx_item(entry) for entry in entries]
+            getattr(self, field).extend(entries)
+        self.files = TaskMatchFilesConfig()
+        return self
+
     @property
     def empty(self) -> bool:
         return not (self.idx or self.ids or self.keys or self.names)