checkpoints_1024 / unit_test.py
ckadirt's picture
Add files using upload-large-folder tool
0a8b79b verified
import numpy as np
import pytest
from utils import filter_and_average_mst, verify_image_patterns, compute_vox_rels, compute_avg_repeat_corrs
# === filter_and_average_mst tests ===
def test_no_mst_images():
vox = np.array([[1,2,3], [4,5,6], [7,8,9]])
vox_image_dict = {0: 'image1.jpg', 1: 'image2.jpg', 2: 'image3.jpg'}
filtered_vox, kept_indices = filter_and_average_mst(vox, vox_image_dict)
np.testing.assert_array_equal(filtered_vox, vox)
np.testing.assert_array_equal(kept_indices, [0, 1, 2])
def test_single_mst_image_set():
vox = np.array([[1,2,3], [4,5,6], [7,8,9], [10,11,12]])
vox_image_dict = {0: 'image1.jpg', 1: 'MST_pairs/image2.jpg', 2: 'image3.jpg', 3: 'MST_pairs/image2.jpg'}
filtered_vox, kept_indices = filter_and_average_mst(vox, vox_image_dict)
expected_vox = np.array([[1,2,3], [7,8,9], [7,8,9]])
expected_indices = [0, 1, 2]
np.testing.assert_array_equal(filtered_vox, expected_vox)
np.testing.assert_array_equal(kept_indices, expected_indices)
def test_multiple_mst_image_sets():
vox = np.array([[1,2,3], [4,5,6], [7,8,9], [7,8,9], [10,11,12], [12,15,12]])
vox_image_dict = {
0: 'image1.jpg',
1: 'MST_pairs/image2.jpg',
2: 'image3.jpg',
3: 'MST_pairs/image2.jpg',
4: 'MST_pairs/image4.jpg',
5: 'MST_pairs/image4.jpg'
}
filtered_vox, kept_indices = filter_and_average_mst(vox, vox_image_dict)
expected_vox = np.array([[1,2,3], [5.5, 6.5, 7.5], [7,8,9], [11,13,12]])
expected_indices = [0, 1, 2, 4]
np.testing.assert_array_equal(filtered_vox, expected_vox)
np.testing.assert_array_equal(kept_indices, expected_indices)
def test_empty_input():
vox = np.array([])
vox_image_dict = {}
filtered_vox, kept_indices = filter_and_average_mst(vox, vox_image_dict)
assert len(filtered_vox) == 0
assert len(kept_indices) == 0
def test_input_shape():
vox = np.random.rand(5, 3)
vox_image_dict = {0: 'a', 1: 'b', 2: 'c', 3: 'd', 4: 'e'}
filtered_vox, _ = filter_and_average_mst(vox, vox_image_dict)
assert filtered_vox.shape[1] == vox.shape[1]
# === verify_image_patterns tests ===
def test_valid_special515():
image_to_indices = {
"all_stimuli/special515/image1.jpg": [[1, 2, 3], []],
"all_stimuli/special515/image2.jpg": [[], [10, 11, 12]],
}
failures = verify_image_patterns(image_to_indices)
assert failures == []
def test_invalid_special515():
image_to_indices = {
"all_stimuli/special515/image1.jpg": [[1, 2], []],
"all_stimuli/special515/image2.jpg": [[1, 2], [3]],
}
failures = verify_image_patterns(image_to_indices)
assert len(failures) == 2
def test_valid_MST_pairs():
image_to_indices = {
"all_stimuli/MST_pairs/image1.png": [[4, 5], [6, 7]],
}
failures = verify_image_patterns(image_to_indices)
assert failures == []
def test_invalid_MST_pairs():
image_to_indices = {
"all_stimuli/MST_pairs/image1.png": [[4, 5, 6], [7]],
}
failures = verify_image_patterns(image_to_indices)
assert len(failures) == 1
def test_valid_other_images():
image_to_indices = {
"all_stimuli/other/image1.png": [[123], []],
"all_stimuli/other/image2.png": [[], [456]],
}
failures = verify_image_patterns(image_to_indices)
assert failures == []
def test_invalid_other_images():
image_to_indices = {
"all_stimuli/other/image1.png": [[123, 124], []],
"all_stimuli/other/image2.png": [[123], [456]],
}
failures = verify_image_patterns(image_to_indices)
assert len(failures) == 2
# === compute_vox_rels tests ===
# def test_reliability_two_repeats():
# np.random.seed(0)
# vox = np.random.rand(70, 10) # 50 trials, 10 voxels
# pairs = [
# [0, 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]
# ]
# rels = compute_vox_rels(vox, pairs, "sub-01", "ses-01")
# assert rels.shape == (10,)
# assert not np.all(np.isnan(rels)), "All voxel reliabilities are NaN!"
# assert np.all((rels >= -1) & (rels <= 1))
# def test_reliability_three_repeats():
# np.random.seed(1)
# vox = np.random.rand(15, 3) # 15 trials, 3 voxels
# pairs = [[0, 1, 2], [3, 4, 5], [6, 7, 8]]
# rels = compute_vox_rels(vox, pairs, "sub-01", "ses-02")
# assert rels.shape == (3,)
# assert not np.all(np.isnan(rels)), "All voxel reliabilities are NaN!"
# assert np.all((rels >= -1) & (rels <= 1))
# def test_reliability_four_repeats_mixed():
# np.random.seed(2)
# vox = np.random.rand(20, 4) # 20 trials, 4 voxels
# pairs = [[0, 1, 2, 3], [4, 5, 6, 7], [8, 9]] # includes 2 and 4 repeats
# rels = compute_vox_rels(vox, pairs, "sub-test", "ses-test")
# assert rels.shape == (4,)
# assert not np.all(np.isnan(rels)), "All voxel reliabilities are NaN!"
# assert np.all((rels >= -1) & (rels <= 1))
# def test_near_uniform_data():
# np.random.seed(42)
# # Add very small noise to a constant baseline
# vox = np.ones((6, 3)) + np.random.normal(0, 1e-5, (6, 3))
# pairs = [[0, 1], [2, 3], [4, 5]]
# rels = compute_vox_rels(vox, pairs, "sub-near-uniform", "ses-01")
# assert rels.shape == (3,)
# assert not np.all(np.isnan(rels)), "All voxel reliabilities are NaN!"
# assert np.all((rels >= -1) & (rels <= 1))
# def test_invalid_pairs_length():
# vox = np.random.rand(10, 3)
# pairs = [[0]] # should raise due to too few repeats
# with pytest.raises(AssertionError):
# compute_vox_rels(vox, pairs, "sub-err", "ses-01")
def test_basic_case():
"""Test with 2 repeats and 2 voxels, with basic correlation"""
vox_repeats = np.random.rand(30, 50)
breakpoint()
rels = compute_avg_repeat_corrs(vox_repeats)
# Expected correlation for each voxel should be the correlation between repeat 0 and repeat 1
assert rels.shape == (2,) # Should return a vector of size 2 (one per voxel)
# Check that the correlation is valid and close to expected value
assert np.all(np.isfinite(rels)) # Ensure no NaNs in the results
for v in range(2): # Check correlation for each voxel
expected_corr = np.corrcoef(vox_repeats[:, v])[0, 1]
assert np.allclose(rels[v], expected_corr, atol=1e-5) # Allow for floating point errors
def test_multiple_repeats():
"""Test with more repeats (3) and multiple voxels (3)"""
vox_repeats = np.array([[1, 2, 3], [2, 3, 4], [3, 4, 5]]) # 3 repeats, 3 voxels
rels = compute_avg_repeat_corrs(vox_repeats)
assert rels.shape == (3,) # Should return a vector of size 3 (one per voxel)
for v in range(3):
assert not np.isnan(rels[v]) # Ensure no NaNs are present
def test_identical_repeats():
"""Test with all identical repeats (perfect correlation)"""
vox_repeats = np.array([[1, 1], [1, 1]]) # Identical repeats, 2 voxels
rels = compute_avg_repeat_corrs(vox_repeats)
assert rels.shape == (2,)
assert np.allclose(rels, 1) # Perfect correlation (should be 1 for all voxels)
def test_anticorrelation():
"""Test with perfect anti-correlation (correlation = -1)"""
vox_repeats = np.array([[1, 2], [2, 1]]) # Perfect anti-correlation between repeats
rels = compute_avg_repeat_corrs(vox_repeats)
assert rels.shape == (2,)
assert np.allclose(rels, -1) # Perfect negative correlation
def test_zero_variance_repeats():
"""Test with repeats having zero variance (e.g., all values are the same)"""
vox_repeats = np.array([[1, 1], [1, 1], [1, 1]]) # Zero variance across repeats
rels = compute_avg_repeat_corrs(vox_repeats)
assert rels.shape == (2,)
# Since variance is zero, the correlation will be NaN
assert np.all(np.isnan(rels))
def test_edge_case_two_repeats_and_one_voxel():
"""Test with only 2 repeats and 1 voxel (minimal edge case)"""
vox_repeats = np.array([[1], [2]]) # 2 repeats, 1 voxel
rels = compute_avg_repeat_corrs(vox_repeats)
assert rels.shape == (1,)
assert np.allclose(rels[0], np.corrcoef([1], [2])[1, 0]) # Correlation between the two repeats