unsloth/tests/saving/test_llm_compressor_install_pin.py
Daniel Han ec4c044e70
Pin llm-compressor auto-install to a vetted version range (#6778)
* Pin llm-compressor auto-install to a vetted version range

install_llm_compressor() auto-installs llm-compressor on first use of an FP8/FP4
compressed export when it is not already present. The install command used the bare
package name, so pip resolved to whatever the configured index served; a compromised,
dependency-confused, or inflated-version ("999.0.0") release could then run under the
Unsloth process at install and import time.

Bound the automatic install to a vetted range
(_LLM_COMPRESSOR_SPEC = "llmcompressor>=0.8.0,<0.13"), which the oneshot /
QuantizationModifier API this uses supports, so pip can no longer jump to an arbitrary
future or inflated version. An already-installed newer llm-compressor is still used
as-is (the import short-circuits), so this only constrains the auto-install, never a
user's own install.

Add UNSLOTH_DISABLE_LLM_COMPRESSOR_AUTOINSTALL=1 to forbid the automatic install
entirely and require a manual, vetted install, for locked-down or air-gapped
environments. Update the manual-install hints to the pinned spec.

Add tests/saving/test_llm_compressor_install_pin.py: static (ast) guards that the spec
stays a bounded pin, that the install command never passes an unpinned llmcompressor
literal, and that the opt-out env gate is evaluated before any install runs.

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

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

* Loosen llm-compressor auto-install ceiling to <1.0 so new models still export

The earlier <0.13 ceiling was too tight: brand-new architectures (for example
Qwen3_5ForConditionalGeneration / qwen3_5, gemma-4 MoE) can require a newer
llm-compressor, and _unsloth_save_compressed_tensors already fails with
"requires a newer llm-compressor" when a scheme is unavailable. Capping the
auto-install at 0.12 would block getting that newer release and break compressed
export for new models.

Widen to llmcompressor>=0.8.0,<1.0. pip still auto-installs the latest 0.x
(where new-architecture support lands), while the <1.0 ceiling continues to block
a jump to an inflated-version ("999.0.0") or 1.0+ dependency-confusion release.
An already-installed newer llm-compressor is still used as-is (the import
short-circuits), and UNSLOTH_DISABLE_LLM_COMPRESSOR_AUTOINSTALL still forbids the
automatic install entirely for locked-down environments.

* Lower llm-compressor floor to 0.6.0 so supported old torch still resolves

The >=0.8.0 floor conflicts with the torch this install pins in its constraints
file. Unsloth supports torch>=2.4, but llm-compressor 0.7.0+ require torch>=2.7
(0.10+ need >=2.9, 0.12+ need >=2.10). On a supported torch 2.4-2.6 box pip then
has no candidate in [0.8.0, 1.0) and FP8/FP4 export fails before quantization.

Lower the floor to 0.6.0 (its metadata only needs torch>=1.7), which never
conflicts with any supported torch. pip still prefers the newest compatible
release, so modern torch continues to get the latest 0.x (0.12.0). The <1.0
ceiling that blocks an inflated-version supply-chain jump is unchanged.

Add a regression test asserting the floor stays <= 0.6.0.

* Cap llm-compressor auto-install ceiling to a vetted minor (<0.13)

A bare <1.0 ceiling still admits any 0.x, so an inflated "0.999.0" served by a
compromised or misconfigured index would win pip's highest-version selection --
the same dependency-confusion this pin is meant to block. Cap the ceiling to the
current vetted minor (<0.13) so that jump is blocked; bump it deliberately, after
vetting, when a newer llm-compressor is needed (e.g. for a brand-new architecture
scheme). Current new models are unaffected: 0.12.0 is < 0.13 and supports them.

The 0.6.0 floor (torch>=1.7 compatible) is unchanged, so resolution still works
across Unsloth's whole supported torch range (2.4 -> 0.6.0 ... 2.12 -> 0.12.0).

Add a regression test asserting the ceiling admits the current vetted release but
blocks an inflated 0.x and the next major.

* Trim comments in the llm-compressor pin (comment-only, no code change)

* Cap llm-compressor auto-install to the exact vetted patch (<=0.12.0)

* [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>
2026-07-01 04:48:38 -07:00

112 lines
4.7 KiB
Python

"""Static guards (no import/network/GPU, like test_save_shell_injection.py) that
install_llm_compressor()'s first-use auto-install of llm-compressor stays version-pinned to a vetted
range and keeps its opt-out env gate, so a compromised/inflated release can't be auto-pulled."""
from __future__ import annotations
import ast
from pathlib import Path
SAVE_PY = Path(__file__).resolve().parents[2] / "unsloth" / "save.py"
_ENV_FLAG = "UNSLOTH_DISABLE_LLM_COMPRESSOR_AUTOINSTALL"
def _module() -> ast.Module:
return ast.parse(SAVE_PY.read_text(encoding = "utf-8"), filename = str(SAVE_PY))
def _get_function(name: str) -> ast.FunctionDef:
for node in ast.walk(_module()):
if isinstance(node, ast.FunctionDef) and node.name == name:
return node
raise AssertionError(f"Function {name} not found in save.py")
def _spec_value():
for node in ast.walk(_module()):
if isinstance(node, ast.Assign) and isinstance(node.value, ast.Constant):
if any(
isinstance(t, ast.Name) and t.id == "_LLM_COMPRESSOR_SPEC" for t in node.targets
):
return node.value.value
return None
def _first_lineno(fn: ast.AST, predicate) -> int | None:
lines = [n.lineno for n in ast.walk(fn) if predicate(n) and hasattr(n, "lineno")]
return min(lines) if lines else None
def test_spec_is_a_bounded_pin() -> None:
spec = _spec_value()
assert spec is not None, "_LLM_COMPRESSOR_SPEC must be defined at module scope"
assert "llmcompressor" in spec, f"spec must name llmcompressor, got {spec!r}"
# A lower and an upper bound: pip cannot jump to an arbitrary (e.g. inflated) future release.
assert ">=" in spec and "<" in spec, f"spec must have lower and upper bounds, got {spec!r}"
def test_ceiling_blocks_inflated_versions() -> None:
"""Cap to the exact vetted patch: block an inflated 0.x, a new major, and any higher in-range patch."""
from packaging.requirements import Requirement
spec = Requirement(_spec_value()).specifier
assert spec.contains("0.12.0"), "the current vetted release must resolve"
assert not spec.contains("0.999.0"), "an inflated 0.x must be blocked"
assert not spec.contains("1.0.0"), "a new major must not be auto-installed"
assert not spec.contains(
"0.12.1"
), "a higher in-range patch must be blocked (cap to the vetted patch)"
assert not spec.contains(
"0.12.999"
), "a crafted higher in-range patch (e.g. on a mirror) must be blocked"
def test_floor_stays_compatible_with_supported_torch() -> None:
"""Floor must stay <=0.6.0: 0.7+ need torch>=2.7, but the pinned torch can be as old as 2.4."""
from packaging.requirements import Requirement
from packaging.version import Version
req = Requirement(_spec_value())
lowers = [Version(s.version) for s in req.specifier if s.operator in (">=", "==", "~=")]
assert lowers, "spec must declare a lower bound"
assert max(lowers) <= Version("0.6.0"), (
f"floor {max(lowers)} requires a torch newer than Unsloth's minimum (2.4); "
"llm-compressor >0.6.0 needs torch>=2.7. Keep the floor <= 0.6.0."
)
def test_install_command_uses_pinned_spec_not_bare_name() -> None:
fn = _get_function("install_llm_compressor")
# No argv list may pass the bare, unpinned package literal "llmcompressor".
for node in ast.walk(fn):
if isinstance(node, ast.List):
for elt in node.elts:
if isinstance(elt, ast.Constant) and elt.value == "llmcompressor":
raise AssertionError(
"install command must not pass an unpinned 'llmcompressor' literal; "
"use the bounded _LLM_COMPRESSOR_SPEC"
)
names = {n.id for n in ast.walk(fn) if isinstance(n, ast.Name)}
assert "_LLM_COMPRESSOR_SPEC" in names, "install command must reference _LLM_COMPRESSOR_SPEC"
def test_optout_env_gate_precedes_subprocess_install() -> None:
fn = _get_function("install_llm_compressor")
env_line = _first_lineno(fn, lambda n: isinstance(n, ast.Constant) and n.value == _ENV_FLAG)
assert env_line is not None, f"{_ENV_FLAG} opt-out must be checked in install_llm_compressor"
def _is_check_call(n: ast.AST) -> bool:
return (
isinstance(n, ast.Call)
and isinstance(n.func, ast.Attribute)
and n.func.attr == "check_call"
and isinstance(n.func.value, ast.Name)
and n.func.value.id == "subprocess"
)
install_line = _first_lineno(fn, _is_check_call)
assert install_line is not None, "expected a subprocess.check_call install in the function"
assert (
env_line < install_line
), "the auto-install opt-out must be evaluated before any package install runs"