61 lines
2.5 KiB
Python
61 lines
2.5 KiB
Python
import os
|
|
import sys
|
|
import unittest
|
|
import torch
|
|
|
|
PROJECT_ROOT = os.path.realpath(os.path.join(os.path.dirname(__file__), "..", ".."))
|
|
MODULE_ROOT = os.path.join(PROJECT_ROOT, "modules")
|
|
sys.path.append(MODULE_ROOT)
|
|
|
|
from pytorch_wrapper.pt_bmm import PtBmm
|
|
|
|
|
|
class TestPtBmm(unittest.TestCase):
|
|
def setUp(self):
|
|
"""Set up test instance."""
|
|
self.node = PtBmm()
|
|
|
|
# Valid test cases: (Tensor A, Tensor B, Expected result)
|
|
self.valid_test_cases = [
|
|
# Case 1: Batch of 2 matrices (3x4) * (4x5) → (2, 3, 5)
|
|
(torch.ones(2, 3, 4), torch.ones(2, 4, 5), torch.full((2, 3, 5), 4.0)),
|
|
|
|
# Case 2: Batch of 3 matrices (2x3) * (3x2) → (3, 2, 2)
|
|
(torch.tensor([[[1.0, 2.0, 3.0], [4.0, 5.0, 6.0]]]),
|
|
torch.tensor([[[7.0, 8.0], [9.0, 10.0], [11.0, 12.0]]]),
|
|
torch.tensor([[[58.0, 64.0], [139.0, 154.0]]])),
|
|
|
|
# Case 3: Identity matrix batch multiplication (should remain the same)
|
|
(torch.eye(3).expand(4, -1, -1), torch.eye(3).expand(4, -1, -1), torch.eye(3).expand(4, -1, -1)),
|
|
]
|
|
|
|
# Invalid test cases (should raise errors)
|
|
self.invalid_test_cases = [
|
|
(torch.ones(3, 4), torch.ones(4, 5)), # 2D tensors (not allowed)
|
|
(torch.ones(2, 3, 4), torch.ones(3, 4, 5)), # Mismatched batch size
|
|
(torch.ones(2, 3, 4, 5), torch.ones(2, 5, 6)), # 4D tensors (not allowed)
|
|
(torch.ones(2, 3, 4), torch.ones(2, 5, 6)), # Matmul shape mismatch within batch
|
|
]
|
|
|
|
def test_batched_matrix_multiplication(self):
|
|
"""Test valid batched matrix multiplication cases."""
|
|
for tens_a, tens_b, expected in self.valid_test_cases:
|
|
with self.subTest(tens_a=tens_a, tens_b=tens_b):
|
|
result, = self.node.f(tens_a, tens_b)
|
|
|
|
# Ensure return value is a tensor
|
|
self.assertTrue(isinstance(result, torch.Tensor), "Output is not a tensor")
|
|
|
|
# Ensure result matches expected batch matrix multiplication
|
|
self.assertTrue(torch.equal(result, expected), f"Expected {expected}, got {result}")
|
|
|
|
def test_invalid_multiplication(self):
|
|
"""Test invalid cases where dimensions do not match."""
|
|
for tens_a, tens_b in self.invalid_test_cases:
|
|
with self.subTest(tens_a=tens_a, tens_b=tens_b):
|
|
with self.assertRaises(Exception, msg=f"torch.bmm() should fail for shapes {tens_a.shape} and {tens_b.shape}"):
|
|
self.node.f(tens_a, tens_b)
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|