Spaces:
Build error
Build error
File size: 8,125 Bytes
80f8f0d 32771a9 80f8f0d f166566 80f8f0d f166566 80f8f0d | 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 | """Tests for Kompress compressor.
Covers:
- Lazy imports: module importable without torch installed
- is_kompress_available(): correct detection of [ml] extra
- KompressConfig / KompressResult: dataclass defaults
- KompressCompressor: passthrough for short content, fallback on error
- Transform interface: apply() method
"""
from unittest.mock import MagicMock, patch
# ── Import safety (the whole point of the fix) ─────────────────────────
class TestLazyImports:
"""The module must be importable without torch/transformers."""
def test_is_kompress_available_importable(self) -> None:
"""is_kompress_available can be imported even without torch."""
from headroom.transforms.kompress_compressor import is_kompress_available
# Should return bool (True or False depending on environment)
result = is_kompress_available()
assert isinstance(result, bool)
def test_module_import_without_torch(self) -> None:
"""Importing the module with torch blocked should not raise."""
import sys
# Block torch AND onnxruntime imports
with patch.dict(
sys.modules,
{"torch": None, "torch.nn": None, "onnxruntime": None},
):
from headroom.transforms.kompress_compressor import (
_is_pytorch_available,
)
# Without both torch and onnxruntime, should return False
assert _is_pytorch_available() is False
# Note: is_kompress_available() may still return True if onnxruntime
# was already imported before patching. Test the individual checkers.
def test_dataclasses_importable_without_torch(self) -> None:
"""KompressConfig, KompressResult, KompressCompressor are importable without torch."""
from headroom.transforms.kompress_compressor import (
KompressCompressor, # noqa: F401
KompressConfig,
KompressResult,
)
# These don't need torch to instantiate
config = KompressConfig()
assert config.device == "auto"
assert config.enable_ccr is True
result = KompressResult(
compressed="hello",
original="hello world",
original_tokens=2,
compressed_tokens=1,
compression_ratio=0.5,
)
assert result.tokens_saved == 1
assert result.savings_percentage == 50.0
# ── KompressResult ──────────────────────────────────────────────────────
class TestKompressResult:
def test_tokens_saved(self) -> None:
from headroom.transforms.kompress_compressor import KompressResult
r = KompressResult(
compressed="a b",
original="a b c d",
original_tokens=4,
compressed_tokens=2,
compression_ratio=0.5,
)
assert r.tokens_saved == 2
def test_tokens_saved_no_negative(self) -> None:
from headroom.transforms.kompress_compressor import KompressResult
r = KompressResult(
compressed="a b c d e",
original="a b c",
original_tokens=3,
compressed_tokens=5,
compression_ratio=1.67,
)
assert r.tokens_saved == 0
def test_savings_percentage_zero_tokens(self) -> None:
from headroom.transforms.kompress_compressor import KompressResult
r = KompressResult(
compressed="",
original="",
original_tokens=0,
compressed_tokens=0,
compression_ratio=1.0,
)
assert r.savings_percentage == 0.0
def test_default_model(self) -> None:
from headroom.transforms.kompress_compressor import HF_MODEL_ID, KompressResult
r = KompressResult(
compressed="x",
original="x y",
original_tokens=2,
compressed_tokens=1,
compression_ratio=0.5,
)
assert r.model_used == HF_MODEL_ID
# ── KompressCompressor (without model) ──────────────────────────────────
class TestKompressCompressorPassthrough:
"""Test compressor behavior that doesn't require the actual model."""
def test_short_content_passthrough(self) -> None:
"""Content under 10 words should pass through unchanged."""
from headroom.transforms.kompress_compressor import KompressCompressor
compressor = KompressCompressor()
result = compressor.compress("hello world")
assert result.compressed == "hello world"
assert result.compression_ratio == 1.0
assert result.original_tokens == 2
assert result.compressed_tokens == 2
def test_empty_content_passthrough(self) -> None:
from headroom.transforms.kompress_compressor import KompressCompressor
compressor = KompressCompressor()
result = compressor.compress("")
assert result.compressed == ""
assert result.compression_ratio == 1.0
def test_fallback_on_model_error(self) -> None:
"""If _load_kompress fails, compress should return passthrough."""
from headroom.transforms.kompress_compressor import KompressCompressor
compressor = KompressCompressor()
long_text = " ".join(f"word{i}" for i in range(20))
with patch(
"headroom.transforms.kompress_compressor._load_kompress",
side_effect=RuntimeError("no model"),
):
result = compressor.compress(long_text)
assert result.compressed == long_text
assert result.compression_ratio == 1.0
# ── Transform interface ─────────────────────────────────────────────────
class TestKompressTransformInterface:
def test_apply_short_messages_unchanged(self) -> None:
"""Messages with <10 words should pass through apply() unchanged."""
from headroom.transforms.kompress_compressor import KompressCompressor
compressor = KompressCompressor()
messages = [
{"role": "user", "content": "hello"},
{"role": "tool", "content": "short"},
]
tokenizer = MagicMock()
tokenizer.count_text = MagicMock(return_value=5)
result = compressor.apply(messages, tokenizer)
assert len(result.messages) == 2
assert result.messages[0]["content"] == "hello"
assert result.messages[1]["content"] == "short"
def test_apply_preserves_user_messages(self) -> None:
"""User messages should never be compressed."""
from headroom.transforms.kompress_compressor import KompressCompressor
compressor = KompressCompressor()
long_text = " ".join(f"word{i}" for i in range(50))
messages = [{"role": "user", "content": long_text}]
tokenizer = MagicMock()
tokenizer.count_text = MagicMock(return_value=50)
with patch(
"headroom.transforms.kompress_compressor._load_kompress",
side_effect=RuntimeError("should not be called"),
):
result = compressor.apply(messages, tokenizer)
assert result.messages[0]["content"] == long_text
# ── unload_kompress_model ───────────────────────────────────────────────
class TestUnloadKompressModel:
def test_unload_when_no_model(self) -> None:
import headroom.transforms.kompress_compressor as kmod
from headroom.transforms.kompress_compressor import unload_kompress_model
# Ensure no model is loaded (previous tests may have set the global)
kmod._kompress_model = None
kmod._kompress_tokenizer = None
# Should return False when no model is loaded
assert unload_kompress_model() is False
|