Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
# SPDX-License-Identifier: AGPL-3.0-only
# Copyright 2026-present the Unsloth AI Inc. team. All rights reserved. See /studio/LICENSE.AGPL-3.0
"""The string-split seed path helpers equal PurePosixPath / Path / a fresh re.search, POSIX and Windows flavours."""
import importlib.util
import random
import re
from pathlib import Path, PurePosixPath, PureWindowsPath
import pytest
def _load(monkeypatch):
pytest.importorskip("fastapi")
pytest.importorskip("structlog")
backend_root = Path(__file__).resolve().parent.parent
monkeypatch.syspath_prepend(str(backend_root))
spec = importlib.util.spec_from_file_location(
"seed_path_split_under_test", backend_root / "routes" / "data_recipe" / "seed.py"
)
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)
return module
_SEGMENTS = [
"data",
"train",
"Train",
"TEST",
"validation",
"dev",
"__hidden",
"__init__",
".git",
".",
"..",
"",
"sub-train",
"en",
"sample-10BT",
"x y",
"a:b",
"C:",
"back\\slash",
"İstanbul",
"é",
"-",
"00",
]
_NAMES = [
"train",
"test",
"train-00000-of-00004",
"questions.train",
"a.csv",
".hidden",
"README",
"x.",
"y..",
"..",
".",
"TRAIN",
"en-train",
"000_00001",
"__train",
"tést",
"dev-1",
"a b",
"c:d",
]
_EXTS = [
".parquet",
".PARQUET",
".json",
".jsonl",
".csv",
".json.gz",
"",
".",
".parquet/",
".csv.parquet",
]
def _paths(count = 3000, seed = 0):
rng = random.Random(seed)
out = [
"",
".",
"..",
"/",
"//",
"///",
"a",
"a/",
"a/.",
"a/..",
"./a",
"/a",
"//a/b",
"a//b",
"a/./b",
"C:/a.parquet",
"C:a.parquet",
"a\\b.parquet",
"a/b.parquet ",
"a/b .parquet",
"a/.parquet",
"a/b.",
"a/b..parquet",
"data/train-00000-of-00001.parquet",
"İ.PARQUET",
"data/x.Parquet",
]
for _ in range(count):
parts = [rng.choice(_SEGMENTS) for _ in range(rng.choice([0, 1, 1, 2, 3]))]
path = "/".join(parts + [rng.choice(_NAMES) + rng.choice(_EXTS)])
if rng.random() < 0.05:
path = "/" + path
if rng.random() < 0.05:
path = path.replace("/", "//", 1)
if rng.random() < 0.03:
path += "/"
out.append(path)
return out
def test_marked_parts_and_loader_facts_match_pure_posix_path(monkeypatch):
seed = _load(monkeypatch)
for path in _paths():
parsed = PurePosixPath(path)
for prefix in (".", "__"):
for which, parts in (("parts", parsed.parts), ("parent_parts", parsed.parent.parts)):
expected = [p for p in parts if p.startswith(prefix) and set(p) != {"."}]
assert seed._marked_parts(path, prefix, which) == expected, (path, prefix, which)
assert seed._loader_path_facts(path)[2] == parsed.name, path
@pytest.mark.parametrize("flavour", [PurePosixPath, PureWindowsPath])
def test_name_parts_match_pathlib(monkeypatch, flavour):
seed = _load(monkeypatch)
monkeypatch.setattr(seed, "Path", flavour)
for path in _paths():
for value in (path, path.lower()):
parsed = flavour(value)
assert seed._name_parts(value) == (parsed.suffix, parsed.stem), (flavour, value)
def test_label_matcher_matches_a_fresh_search(monkeypatch):
seed = _load(monkeypatch)
labels = ["train", "dev", "en", "sample-10bt", "a.b", "x+y", "[", "", "validation"]
for path in _paths(500):
name = path.lower()
for label in labels:
expected = (
re.search(rf"(?:^|{seed._SEP}){re.escape(label)}(?:$|{seed._SEP})", name)
is not None
)
assert seed._label_in_name(name, label) == expected, (name, label)
@pytest.mark.parametrize("flavour", [PurePosixPath, PureWindowsPath])
def test_split_rank_and_suffix_filter_match_pathlib(monkeypatch, flavour):
seed = _load(monkeypatch)
monkeypatch.setattr(seed, "Path", flavour)
paths = _paths(1500, seed = 1)
for path in paths:
lowered = path.lower()
labels = seed._split_labels("train")
if seed._split_folder(lowered, labels):
expected = 0
elif any(seed._label_in_name(flavour(lowered).stem, label) for label in labels):
expected = 1
else:
expected = 2
assert seed._split_rank(path, "train") == expected, (flavour, path)
for suffix in (".parquet", ".json", ".JSONL", ".csv", ""):
group = seed._builder_exts(suffix)
expected = [p for p in paths if flavour(p).suffix.lower() in group]
assert seed._of_suffix(paths, suffix) == expected, (flavour, suffix)