Skip to content

Commit

Permalink
Converted the remaining CRLF line endings to LF
Browse files Browse the repository at this point in the history
  • Loading branch information
TakeshiMusgrave committed Dec 20, 2020
1 parent 15050e0 commit 025e647
Show file tree
Hide file tree
Showing 4 changed files with 681 additions and 681 deletions.
274 changes: 137 additions & 137 deletions tests/miners/test_batch_hard_miner.py
View file
Original file line number Diff line number Diff line change
@@ -1,137 +1,137 @@
import unittest
from .. import TEST_DTYPES, TEST_DEVICE
import torch
from pytorch_metric_learning.miners import BatchHardMiner
from pytorch_metric_learning.utils import common_functions as c_f
import numpy as np
from pytorch_metric_learning.distances import CosineSimilarity, LpDistance


class TestBatchHardMiner(unittest.TestCase):
@classmethod
def setUpClass(self):
self.dist_miner = BatchHardMiner(
distance=LpDistance(normalize_embeddings=False)
)
self.normalized_dist_miner = BatchHardMiner(
distance=LpDistance(normalize_embeddings=True)
)
self.normalized_dist_miner_squared = BatchHardMiner(
distance=LpDistance(normalize_embeddings=True, power=2)
)
self.sim_miner = BatchHardMiner(distance=CosineSimilarity())
self.labels = torch.LongTensor([0, 0, 1, 1, 0, 2, 1, 1, 1])
self.correct_a = torch.LongTensor([0, 1, 2, 3, 4, 6, 7, 8]).to(TEST_DEVICE)
self.correct_p = torch.LongTensor([4, 4, 8, 8, 0, 2, 2, 2]).to(TEST_DEVICE)
self.correct_n = [
torch.LongTensor([2, 2, 1, 4, 3, 5, 5, 5]).to(TEST_DEVICE),
torch.LongTensor([2, 2, 1, 4, 5, 5, 5, 5]).to(TEST_DEVICE),
]

@classmethod
def tearDown(self):
torch.cuda.empty_cache()

def test_dist_mining(self):
for dtype in TEST_DTYPES:
embeddings = torch.arange(9).type(dtype).unsqueeze(1).to(TEST_DEVICE)
a, p, n = self.dist_miner(embeddings, self.labels)
self.helper(a, p, n)
self.assertTrue(self.dist_miner.hardest_pos_pair == 6)
self.assertTrue(self.dist_miner.hardest_neg_pair == 1)

def test_normalized_dist_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.normalized_dist_miner(embeddings, self.labels)
self.helper(a, p, n)
correct_hardest_pos_pair = torch.sqrt(
torch.sum((embeddings[2] - embeddings[8]) ** 2)
).item()
correct_hardest_neg_pair = torch.sqrt(
torch.sum((embeddings[1] - embeddings[2]) ** 2)
).item()
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.normalized_dist_miner.hardest_pos_pair,
correct_hardest_pos_pair,
places=places,
)
self.assertAlmostEqual(
self.normalized_dist_miner.hardest_neg_pair,
correct_hardest_neg_pair,
places=places,
)

def test_normalized_dist_squared_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.normalized_dist_miner_squared(embeddings, self.labels)
self.helper(a, p, n)
correct_hardest_pos_pair = torch.sum(
(embeddings[2] - embeddings[8]) ** 2
).item()
correct_hardest_neg_pair = torch.sum(
(embeddings[1] - embeddings[2]) ** 2
).item()
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.normalized_dist_miner_squared.hardest_pos_pair,
correct_hardest_pos_pair,
places=places,
)
self.assertAlmostEqual(
self.normalized_dist_miner_squared.hardest_neg_pair,
correct_hardest_neg_pair,
places=places,
)

def test_sim_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.sim_miner(embeddings, self.labels)
self.helper(a, p, n)
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.sim_miner.hardest_pos_pair,
np.cos(np.radians(120)),
places=places,
)
self.assertAlmostEqual(
self.sim_miner.hardest_neg_pair,
np.cos(np.radians(20)),
places=places,
)

def helper(self, a, p, n):
self.assertTrue(torch.equal(a, self.correct_a))
self.assertTrue(torch.equal(p, self.correct_p))
self.assertTrue(any(torch.equal(n, cn) for cn in self.correct_n))

def test_empty_output(self):
batch_size = 32
for dtype in TEST_DTYPES:
embeddings = torch.randn(batch_size, 64).type(dtype).to(TEST_DEVICE)
labels = torch.arange(batch_size)
for miner in [
self.dist_miner,
self.normalized_dist_miner,
self.normalized_dist_miner_squared,
self.sim_miner,
]:
a, p, n = miner(embeddings, labels)
self.assertTrue(len(a) == 0)
self.assertTrue(len(p) == 0)
self.assertTrue(len(n) == 0)
self.assertTrue(miner.hardest_pos_pair == 0)
self.assertTrue(miner.hardest_neg_pair == 0)
self.assertTrue(miner.hardest_triplet == 0)
import unittest
from .. import TEST_DTYPES, TEST_DEVICE
import torch
from pytorch_metric_learning.miners import BatchHardMiner
from pytorch_metric_learning.utils import common_functions as c_f
import numpy as np
from pytorch_metric_learning.distances import CosineSimilarity, LpDistance


class TestBatchHardMiner(unittest.TestCase):
@classmethod
def setUpClass(self):
self.dist_miner = BatchHardMiner(
distance=LpDistance(normalize_embeddings=False)
)
self.normalized_dist_miner = BatchHardMiner(
distance=LpDistance(normalize_embeddings=True)
)
self.normalized_dist_miner_squared = BatchHardMiner(
distance=LpDistance(normalize_embeddings=True, power=2)
)
self.sim_miner = BatchHardMiner(distance=CosineSimilarity())
self.labels = torch.LongTensor([0, 0, 1, 1, 0, 2, 1, 1, 1])
self.correct_a = torch.LongTensor([0, 1, 2, 3, 4, 6, 7, 8]).to(TEST_DEVICE)
self.correct_p = torch.LongTensor([4, 4, 8, 8, 0, 2, 2, 2]).to(TEST_DEVICE)
self.correct_n = [
torch.LongTensor([2, 2, 1, 4, 3, 5, 5, 5]).to(TEST_DEVICE),
torch.LongTensor([2, 2, 1, 4, 5, 5, 5, 5]).to(TEST_DEVICE),
]

@classmethod
def tearDown(self):
torch.cuda.empty_cache()

def test_dist_mining(self):
for dtype in TEST_DTYPES:
embeddings = torch.arange(9).type(dtype).unsqueeze(1).to(TEST_DEVICE)
a, p, n = self.dist_miner(embeddings, self.labels)
self.helper(a, p, n)
self.assertTrue(self.dist_miner.hardest_pos_pair == 6)
self.assertTrue(self.dist_miner.hardest_neg_pair == 1)

def test_normalized_dist_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.normalized_dist_miner(embeddings, self.labels)
self.helper(a, p, n)
correct_hardest_pos_pair = torch.sqrt(
torch.sum((embeddings[2] - embeddings[8]) ** 2)
).item()
correct_hardest_neg_pair = torch.sqrt(
torch.sum((embeddings[1] - embeddings[2]) ** 2)
).item()
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.normalized_dist_miner.hardest_pos_pair,
correct_hardest_pos_pair,
places=places,
)
self.assertAlmostEqual(
self.normalized_dist_miner.hardest_neg_pair,
correct_hardest_neg_pair,
places=places,
)

def test_normalized_dist_squared_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.normalized_dist_miner_squared(embeddings, self.labels)
self.helper(a, p, n)
correct_hardest_pos_pair = torch.sum(
(embeddings[2] - embeddings[8]) ** 2
).item()
correct_hardest_neg_pair = torch.sum(
(embeddings[1] - embeddings[2]) ** 2
).item()
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.normalized_dist_miner_squared.hardest_pos_pair,
correct_hardest_pos_pair,
places=places,
)
self.assertAlmostEqual(
self.normalized_dist_miner_squared.hardest_neg_pair,
correct_hardest_neg_pair,
places=places,
)

def test_sim_mining(self):
for dtype in TEST_DTYPES:
angles = [0, 20, 40, 60, 80, 100, 120, 140, 160]
embeddings = torch.tensor(
[c_f.angle_to_coord(a) for a in angles], dtype=dtype
).to(TEST_DEVICE)
a, p, n = self.sim_miner(embeddings, self.labels)
self.helper(a, p, n)
places = 2 if dtype == torch.float16 else 5
self.assertAlmostEqual(
self.sim_miner.hardest_pos_pair,
np.cos(np.radians(120)),
places=places,
)
self.assertAlmostEqual(
self.sim_miner.hardest_neg_pair,
np.cos(np.radians(20)),
places=places,
)

def helper(self, a, p, n):
self.assertTrue(torch.equal(a, self.correct_a))
self.assertTrue(torch.equal(p, self.correct_p))
self.assertTrue(any(torch.equal(n, cn) for cn in self.correct_n))

def test_empty_output(self):
batch_size = 32
for dtype in TEST_DTYPES:
embeddings = torch.randn(batch_size, 64).type(dtype).to(TEST_DEVICE)
labels = torch.arange(batch_size)
for miner in [
self.dist_miner,
self.normalized_dist_miner,
self.normalized_dist_miner_squared,
self.sim_miner,
]:
a, p, n = miner(embeddings, labels)
self.assertTrue(len(a) == 0)
self.assertTrue(len(p) == 0)
self.assertTrue(len(n) == 0)
self.assertTrue(miner.hardest_pos_pair == 0)
self.assertTrue(miner.hardest_neg_pair == 0)
self.assertTrue(miner.hardest_triplet == 0)
Loading

0 comments on commit 025e647

Please sign in to comment.