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

Studio: make code comments and docstrings more succinct (#6029)

Browse source 
Trim and tighten code comments and docstrings across studio/ Python. Comment-only: every changed file verified code-identical to main via AST/token comparison.
This commit is contained in:
Daniel Han 2026-06-08 23:07:28 -07:00 • committed by GitHub
commit 8292e699e4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
205 changed files with 8065 additions and 8912 deletions
Download patch file Download diff file Expand all files Collapse all files

12
studio/backend/tests/test_llama_cpp_start_failure_classification.py
View file

@ -4,10 +4,10 @@
"""Tests for LlamaCppBackend._classify_llama_start_failure.
When llama-server exits before becoming healthy, load_model turns its
captured stdout/stderr into a user-facing reason. A diffusion / image
GGUF (FLUX, Qwen-Image, ...) is a valid file with plenty of memory, so
the generic "invalid file or out of memory" message is actively
misleading (issue #5842). These tests pin the classification.
captured stdout/stderr into a user-facing reason. A diffusion/image GGUF
(FLUX, Qwen-Image, ...) is a valid file with plenty of memory, so the
generic "invalid file or out of memory" message is misleading (issue
#5842). These tests pin the classification.
"""
from __future__ import annotations
@ -22,8 +22,8 @@ _BACKEND_DIR = str(Path(__file__).resolve().parent.parent)
if _BACKEND_DIR not in sys.path:
sys.path.insert(0, _BACKEND_DIR)
# Match the stubbing pattern in sibling tests so the module imports in a
# lightweight env without fastapi.
# Match sibling tests' stubbing so the module imports in a lightweight
# env without fastapi.
_loggers_stub = _types.ModuleType("loggers")
_loggers_stub.get_logger = lambda name: __import__("logging").getLogger(name)
sys.modules.setdefault("loggers", _loggers_stub)