File size: 9,053 Bytes
c34ff1f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 | """Integration tests for launch_vllm.py provenance artifacts.
Includes a venv-isolation test that verifies the script imports successfully
even when `speculators` is not on sys.path — reproducing the exact failure
mode that caused PR #958 to be reverted in PR #1008.
"""
from __future__ import annotations
import os
import subprocess
import sys
from pathlib import Path
import pytest
SCRIPTS_DIR = Path(__file__).resolve().parents[3] / "scripts"
LAUNCH_VLLM = str(SCRIPTS_DIR / "launch_vllm.py")
MODEL = "gpt2"
DRAFTER = "some-org/drafter-model"
@pytest.fixture
def provenance_dir(tmp_path: Path) -> Path:
return tmp_path / "provenance"
class TestVenvIsolation:
"""launch_vllm.py must import without speculators installed (vLLM venv)."""
def test_imports_without_speculators_on_path(self):
"""No ImportError when speculators is absent — the root cause of #1008."""
# Strip every path entry that could resolve `speculators` (site/dist
# packages, cwd, its own source tree), then *assert* it's unresolvable
# via find_spec before importing — a robust guard that fails loudly if
# the stripping was incomplete, rather than silently passing.
snippet = (
"import importlib.util, sys, sysconfig; "
"paths = sysconfig.get_paths(); "
"bad = {paths['purelib'], paths['platlib'], ''}; "
"sys.path = [p for p in sys.path if p and p not in bad "
"and 'site-packages' not in p and 'dist-packages' not in p "
"and 'speculators' not in p]; "
f"sys.path.insert(0, {str(SCRIPTS_DIR)!r}); "
"assert importlib.util.find_spec('speculators') is None, "
"('speculators still resolvable: %r' % sys.path); "
"import launch_vllm"
)
env = {k: v for k, v in os.environ.items() if k != "PYTHONPATH"}
result = subprocess.run( # noqa: S603
[sys.executable, "-c", snippet],
capture_output=True,
text=True,
timeout=30,
check=False,
env=env,
)
assert result.returncode == 0, (
f"launch_vllm.py failed to import when speculators was absent.\n"
f"stderr: {result.stderr}"
)
class TestLaunchVllmProvenance:
"""Train subcommand integration tests."""
def _run(
self, provenance_dir: Path, *extra_args: str
) -> subprocess.CompletedProcess:
return subprocess.run( # noqa: S603
[
sys.executable,
LAUNCH_VLLM,
"train",
MODEL,
"--dry-run",
"--provenance-dir",
str(provenance_dir),
*extra_args,
],
capture_output=True,
text=True,
timeout=120,
check=False,
)
def test_creates_all_provenance_files(self, provenance_dir: Path):
result = self._run(provenance_dir)
assert result.returncode == 0, result.stderr
assert (provenance_dir / "vllm_command.txt").exists()
assert (provenance_dir / "vllm.patch").exists()
assert (provenance_dir / "checkpoint_sha256.txt").exists()
def test_vllm_command_contains_metadata(self, provenance_dir: Path):
self._run(provenance_dir)
content = (provenance_dir / "vllm_command.txt").read_text()
assert "# Timestamp:" in content
assert "# Python:" in content
assert "# Git SHA:" in content
assert "# vllm:" in content
assert "vllm.entrypoints" in content
def test_checkpoint_sha256_remote_model(self, provenance_dir: Path):
self._run(provenance_dir)
content = (provenance_dir / "checkpoint_sha256.txt").read_text()
assert "not a local path" in content
def test_no_provenance_written_without_flag(self, tmp_path: Path):
"""Provenance is opt-in: no files/dirs created unless --provenance-dir."""
result = subprocess.run( # noqa: S603
[sys.executable, LAUNCH_VLLM, "train", MODEL, "--dry-run"],
capture_output=True,
text=True,
cwd=str(tmp_path),
timeout=120,
check=False,
)
assert result.returncode == 0, result.stderr
# Nothing should be written to the working directory.
assert list(tmp_path.iterdir()) == []
def test_implicit_train_subcommand(self, provenance_dir: Path):
"""MODEL without 'train' prefix defaults to train mode."""
result = subprocess.run( # noqa: S603
[
sys.executable,
LAUNCH_VLLM,
MODEL,
"--dry-run",
"--provenance-dir",
str(provenance_dir),
],
capture_output=True,
text=True,
timeout=120,
check=False,
)
assert result.returncode == 0, result.stderr
content = (provenance_dir / "vllm_command.txt").read_text()
assert "extract_hidden_states" in content
def test_flags_only_falls_back_to_train(self):
"""Unknown flags without explicit subcommand fall back to train."""
result = subprocess.run( # noqa: S603
[sys.executable, LAUNCH_VLLM, "--some-vllm-flag"],
capture_output=True,
text=True,
timeout=120,
check=False,
)
assert result.returncode != 0
assert "model" in result.stderr.lower()
def test_no_args_shows_usage_error(self):
"""Bare launch_vllm.py with no arguments reports missing model."""
result = subprocess.run( # noqa: S603
[sys.executable, LAUNCH_VLLM],
capture_output=True,
text=True,
timeout=120,
check=False,
)
assert result.returncode != 0
assert "model" in result.stderr.lower()
class TestLaunchVllmEvalMode:
"""Eval subcommand integration tests."""
def _run(
self, provenance_dir: Path, *extra_args: str
) -> subprocess.CompletedProcess:
return subprocess.run( # noqa: S603
[
sys.executable,
LAUNCH_VLLM,
"eval",
MODEL,
"--spec-model",
DRAFTER,
"--spec-tokens",
"3",
"--dry-run",
"--provenance-dir",
str(provenance_dir),
*extra_args,
],
capture_output=True,
text=True,
timeout=120,
check=False,
)
def test_eval_mode_creates_provenance(self, provenance_dir: Path):
result = self._run(provenance_dir)
assert result.returncode == 0, result.stderr
assert (provenance_dir / "vllm_command.txt").exists()
assert (provenance_dir / "vllm.patch").exists()
assert (provenance_dir / "checkpoint_sha256.txt").exists()
assert (provenance_dir / "drafter_checkpoint_sha256.txt").exists()
def test_eval_command_contains_spec_flags(self, provenance_dir: Path):
self._run(provenance_dir)
content = (provenance_dir / "vllm_command.txt").read_text()
assert "--spec-model" in content
assert DRAFTER in content
assert "--spec-tokens" in content
assert "3" in content
def test_eval_command_no_train_flags(self, provenance_dir: Path):
self._run(provenance_dir)
content = (provenance_dir / "vllm_command.txt").read_text()
assert "--speculative_config" not in content
assert "--kv_transfer_config" not in content
assert "extract_hidden_states" not in content
def test_eval_mode_requires_spec_model(self, tmp_path: Path):
result = subprocess.run( # noqa: S603
[
sys.executable,
LAUNCH_VLLM,
"eval",
MODEL,
"--dry-run",
"--provenance-dir",
str(tmp_path / "prov"),
],
capture_output=True,
text=True,
timeout=120,
check=False,
)
assert result.returncode != 0
assert "--spec-model" in result.stderr
def test_eval_mode_with_spec_method(self, provenance_dir: Path):
result = self._run(provenance_dir, "--spec-method", "dflash")
assert result.returncode == 0, result.stderr
content = (provenance_dir / "vllm_command.txt").read_text()
assert "--spec-method" in content
assert "dflash" in content
def test_eval_mode_passes_extra_vllm_args(self, provenance_dir: Path):
result = self._run(provenance_dir, "--", "--port", "8080", "--tp", "2")
assert result.returncode == 0, result.stderr
content = (provenance_dir / "vllm_command.txt").read_text()
assert "--port" in content
assert "8080" in content
assert "--tp" in content
|