ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Reduce and tighten comments and docstrings across the test suite (#6429)

Browse source 
* Reduce and tighten comments and docstrings in tests

Shorten verbose comments and docstrings across the test suite without
changing any test logic. Remove narration that restates the next line,
collapse long module and test docstrings to a single line, and drop banner
separators. Keep regression context (issue and PR references, run ids),
skip reasons, mocking and timing rationale, license headers, lint and type
directives, and commented-out code.

Comments and docstrings only: an AST signature check confirms no code,
assertions, or string literals changed, and the suite byte-compiles cleanly.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
This commit is contained in:
Daniel Han 2026-06-18 01:07:09 -07:00 • committed by GitHub
commit a6dc10dad2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
122 changed files with 1847 additions and 4511 deletions
Download patch file Download diff file Expand all files Collapse all files

26
tests/python/test_construct_chat_template_validation.py
View file

@ -1,15 +1,10 @@
"""Negative-path validation tests for unsloth.chat_templates.construct_chat_template.
Regression coverage for the str.find() / regex no-match guards added in
PR #5763 follow-up: missing placeholders or unrecoverable two-example
structures must raise RuntimeError with a clear message, not IndexError
or AttributeError, and must never silently drop the last character via
s[:-1].
Uses a minimal fake tokenizer so the cases run on CPU-only CI without
HF_TOKEN and without downloading a gated model. The validation paths
exercised here fail before construct_chat_template reaches any heavy
tokenizer interaction, so the stub stays small.
Regression coverage for the no-match guards added in the PR #5763 follow-up:
missing placeholders or unrecoverable two-example structures must raise
RuntimeError with a clear message (not IndexError/AttributeError) and must
not silently drop the last char via s[:-1]. A minimal fake tokenizer keeps
the cases CPU-only (no HF_TOKEN, no gated download).
"""
import pytest
@ -18,8 +13,7 @@ from unsloth.chat_templates import construct_chat_template
class _FakeTokenizer:
"""Minimum surface construct_chat_template touches before the
validation guards fire."""
"""Minimal surface construct_chat_template touches before the guards fire."""
name_or_path = "fake/tokenizer"
eos_token = "</s>"
@ -48,9 +42,8 @@ def test_missing_placeholder_in_chat_template_raises(template, expected_in_messa
def test_single_pair_template_raises_clear_error_not_attribute_error():
"""One {INPUT}/{OUTPUT} pair (rather than the required two) used to
crash with AttributeError on `found.group(1)` after the for-loop
broke without setting `found`. Must raise RuntimeError now."""
"""A single {INPUT}/{OUTPUT} pair must raise RuntimeError, not the old
AttributeError on `found.group(1)` when the loop broke without setting `found`."""
template = "user: {INPUT}\nassistant: {OUTPUT}\n"
with pytest.raises(RuntimeError):
construct_chat_template(
@ -71,7 +64,6 @@ def test_error_message_excerpt_is_bounded():
extra_eos_tokens = ["</s>"],
)
msg = str(exc_info.value)
# Excerpt is repr-quoted and capped; total message should stay well
# under the template length.
# Excerpt is capped well under the template length.
assert len(msg) < 1000
assert "{OUTPUT}" in msg