Spaces:
Sleeping
Sleeping
File size: 9,193 Bytes
5caedb4 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 |
from unittest.mock import MagicMock
import numpy as np
import pandas as pd
import pytest
from scipy.special import softmax
from sklearn.metrics import log_loss, roc_auc_score
from llm_studio.src.metrics.text_causal_classification_modeling_metrics import (
accuracy_score,
auc_score,
logloss_score,
)
@pytest.fixture
def mock_val_df():
return pd.DataFrame()
def test_accuracy_score_binary_perfect_match(mock_val_df):
results = {
"predictions": [[1], [0], [1], [0]],
"target_text": ["1", "0", "1", "0"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([1.0, 1.0, 1.0, 1.0]))
def test_accuracy_score_binary_no_match(mock_val_df):
results = {
"predictions": [[1], [1], [1], [1]],
"target_text": ["0", "0", "0", "0"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([0.0, 0.0, 0.0, 0.0]))
def test_accuracy_score_binary_mixed_results(mock_val_df):
results = {
"predictions": [[1], [0], [1], [0]],
"target_text": ["1", "1", "0", "0"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([1.0, 0.0, 0.0, 1.0]))
def test_accuracy_score_multiclass_perfect_match(mock_val_df):
results = {
"predictions": [[0], [1], [2], [3], [4]],
"target_text": ["0", "1", "2", "3", "4"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([1.0, 1.0, 1.0, 1.0, 1.0]))
def test_accuracy_score_multiclass_no_match(mock_val_df):
results = {
"predictions": [[1], [2], [3], [4], [0]],
"target_text": ["0", "1", "2", "3", "4"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([0.0, 0.0, 0.0, 0.0, 0.0]))
def test_accuracy_score_multiclass_mixed_results(mock_val_df):
results = {
"predictions": [[0], [1], [2], [2], [4]],
"target_text": ["0", "1", "2", "3", "3"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([1.0, 1.0, 1.0, 0.0, 0.0]))
def test_accuracy_score_invalid_input_empty(mock_val_df):
results = {"predictions": [], "target_text": []}
cfg = MagicMock()
with pytest.raises(ValueError):
accuracy_score(cfg, results, mock_val_df)
def test_accuracy_score_invalid_input_unequal_length(mock_val_df):
results = {"predictions": [[1], [0]], "target_text": ["1", "0", "2"]}
cfg = MagicMock()
with pytest.raises(ValueError):
accuracy_score(cfg, results, mock_val_df)
def test_accuracy_score_ignore_raw_results(mock_val_df):
results = {"predictions": [[1], [0], [2]], "target_text": ["1", "1", "2"]}
cfg = MagicMock()
raw_results = True
score = accuracy_score(cfg, results, mock_val_df, raw_results)
assert np.array_equal(score, np.array([1.0, 0.0, 1.0]))
def test_accuracy_score_large_class_numbers(mock_val_df):
results = {
"predictions": [[10], [20], [30], [40], [50]],
"target_text": ["10", "20", "30", "40", "60"],
}
cfg = MagicMock()
score = accuracy_score(cfg, results, mock_val_df)
assert np.array_equal(score, np.array([1.0, 1.0, 1.0, 1.0, 0.0]))
def test_auc_score_binary_classification(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {
"logits": [[0.1, 0.9], [0.8, 0.2], [0.3, 0.7], [0.9, 0.1]],
"target_text": ["1", "0", "1", "0"],
}
score = auc_score(cfg, results, mock_val_df)
expected_score = roc_auc_score([1, 0, 1, 0], [0.9, 0.2, 0.7, 0.1])
assert np.isclose(score, expected_score)
def test_auc_score_multiclass_classification(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 3
results = {
"logits": [[0.1, 0.8, 0.1], [0.7, 0.2, 0.1], [0.1, 0.1, 0.8], [0.3, 0.3, 0.4]],
"target_text": ["1", "0", "2", "2"],
}
score = auc_score(cfg, results, mock_val_df)
expected_score = roc_auc_score(
np.eye(3)[[1, 0, 2, 2]], np.array(results["logits"]), multi_class="ovr"
)
assert np.allclose(score, expected_score)
def test_auc_score_invalid_input_empty(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {"logits": [], "target_text": []}
with pytest.raises(ValueError):
auc_score(cfg, results, mock_val_df)
def test_auc_score_invalid_input_unequal_length(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {"logits": [[0.1, 0.9], [0.8, 0.2]], "target_text": ["1", "2", "0", "2"]}
with pytest.raises(ValueError):
auc_score(cfg, results, mock_val_df)
def test_auc_score_ignore_val_df_and_raw_results(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {"logits": [[0.1, 0.9], [0.8, 0.2]], "target_text": ["1", "0"]}
raw_results = True
score = auc_score(cfg, results, "This should be ignored", raw_results)
expected_score = roc_auc_score([1, 0], [0.9, 0.2])
assert np.isclose(score, expected_score)
def test_auc_score_different_number_of_classes(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 4
results = {
"logits": [
[0.1, 0.7, 0.1, 0.1],
[0.6, 0.2, 0.1, 0.1],
[0.1, 0.1, 0.7, 0.1],
[0.2, 0.2, 0.3, 0.3],
],
"target_text": ["1", "0", "2", "3"],
}
score = auc_score(cfg, results, mock_val_df)
expected_score = roc_auc_score(
np.eye(4)[[1, 0, 2, 3]], np.array(results["logits"]), multi_class="ovr"
)
assert np.allclose(score, expected_score)
def test_logloss_score_binary_classification(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
cfg.dataset.answer_column = ["label"]
results = {
"probabilities": softmax(
[[0.1, 0.9], [0.8, 0.2], [0.3, 0.7], [0.9, 0.1]], axis=1
),
"target_text": ["1", "0", "1", "0"],
}
score = logloss_score(cfg, results, mock_val_df)
expected_score = log_loss([1, 0, 1, 0], results["probabilities"])
assert np.isclose(score, expected_score)
def test_logloss_score_multiclass_classification(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 3
cfg.dataset.answer_column = ["label"]
results = {
"probabilities": softmax(
[[0.1, 0.8, 0.1], [0.7, 0.2, 0.1], [0.1, 0.1, 0.8], [0.3, 0.3, 0.4]], axis=1
),
"target_text": ["1", "0", "2", "2"],
}
score = logloss_score(cfg, results, mock_val_df)
expected_score = log_loss(np.eye(3)[[1, 0, 2, 2]], results["probabilities"])
assert np.isclose(score, expected_score)
def test_logloss_score_multilabel_classification(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 3
cfg.dataset.answer_column = ["label1", "label2", "label3"]
results = {
"probabilities": [
[0.1, 0.8, 0.1],
[0.7, 0.2, 0.1],
[0.1, 0.1, 0.8],
[0.3, 0.3, 0.4],
],
"target_text": ["1,0,1", "0,1,0", "1,1,0", "0,0,1"],
}
score = logloss_score(cfg, results, mock_val_df)
expected_scores = []
for i in range(3):
expected_scores.append(
log_loss(
[int(t.split(",")[i]) for t in results["target_text"]],
[p[i] for p in results["probabilities"]],
)
)
expected_score = np.mean(expected_scores)
assert np.isclose(score, expected_score)
def test_logloss_score_invalid_input_empty(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {"probabilities": [], "target_text": []}
with pytest.raises(ValueError):
logloss_score(cfg, results, mock_val_df)
def test_logloss_score_invalid_input_unequal_length(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
results = {
"probabilities": [[0.1, 0.9], [0.8, 0.2]],
"target_text": ["1", "2", "0"],
}
with pytest.raises(ValueError):
logloss_score(cfg, results, mock_val_df)
def test_logloss_score_ignore_val_df_and_raw_results(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
cfg.dataset.answer_column = ["label"]
results = {"probabilities": [[0.1, 0.9], [0.8, 0.2]], "target_text": ["1", "0"]}
raw_results = True
score = logloss_score(cfg, results, "This should be ignored", raw_results)
expected_score = log_loss([1, 0], results["probabilities"])
assert np.isclose(score, expected_score)
def test_logloss_score_extreme_probabilities(mock_val_df):
cfg = MagicMock()
cfg.dataset.num_classes = 2
cfg.dataset.answer_column = ["label"]
results = {
"probabilities": [[0.0001, 0.9999], [0.9999, 0.0001]],
"target_text": ["1", "0"],
}
score = logloss_score(cfg, results, mock_val_df)
expected_score = log_loss([1, 0], results["probabilities"])
assert np.isclose(score, expected_score)
|