import torch
import torch_npu
from torch_npu.testing.testcase import TestCase, run_tests
class TestHammingWindow(TestCase):
def cpu_op_exec(self, window_length, dtype):
output = torch.hamming_window(window_length, dtype=dtype)
output = output.numpy()
return output
def npu_op_exec(self, window_length, dtype):
output = torch.hamming_window(window_length, dtype=dtype, device='npu')
output = output.to('cpu')
output = output.numpy()
return output
def cpu_op_exec_periodic(self, window_length, periodic, dtype):
output = torch.hamming_window(window_length, periodic, dtype=dtype)
output = output.numpy()
return output
def npu_op_exec_periodic(self, window_length, periodic, dtype):
output = torch.hamming_window(window_length, periodic, dtype=dtype, device='npu')
output = output.to('cpu')
output = output.numpy()
return output
def cpu_op_exec_periodic_alpha(self, window_length, periodic, alpha, dtype):
output = torch.hamming_window(window_length, periodic, alpha, dtype=dtype)
output = output.numpy()
return output
def npu_op_exec_periodic_alpha(self, window_length, periodic, alpha, dtype):
output = torch.hamming_window(window_length, periodic, alpha, dtype=dtype, device='npu')
output = output.to('cpu')
output = output.numpy()
return output
def cpu_op_exec_periodic_alpha_beta(self, window_length, periodic, alpha, beta, dtype):
output = torch.hamming_window(window_length, periodic, alpha, beta, dtype=dtype)
output = output.numpy()
return output
def npu_op_exec_periodic_alpha_beta(self, window_length, periodic, alpha, beta, dtype):
output = torch.hamming_window(window_length, periodic, alpha, beta, dtype=dtype, device='npu')
output = output.to('cpu')
output = output.numpy()
return output
def test_hamming_window(self):
shape_format = [
[0, torch.float32],
[1, torch.float32],
[7, torch.float32],
[12, torch.float32]]
for item in shape_format:
cpu_output = self.cpu_op_exec(*item)
npu_output = self.npu_op_exec(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hamming_window_periodic(self):
shape_format = [
[0, False, torch.float32],
[1, False, torch.float32],
[7, False, torch.float32],
[12, False, torch.float32]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic(*item)
npu_output = self.npu_op_exec_periodic(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hamming_window_periodic_alpha(self):
shape_format = [
[0, True, 0.22, torch.float32],
[0, True, 2.2, torch.float32],
[1, True, 0.22, torch.float32],
[1, True, 2.0, torch.float32],
[7, True, 0.22, torch.float32],
[7, True, 2.0, torch.float32],
[12, True, 0.22, torch.float32],
[12, True, 2.0, torch.float32],
[0, False, 0.22, torch.float32],
[0, False, 2.2, torch.float32],
[1, False, 2.0, torch.float32],
[7, False, 2.0, torch.float32],
[12, False, 1.1, torch.float32]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic_alpha(*item)
npu_output = self.npu_op_exec_periodic_alpha(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hammingwindow_periodic_alpha_beta(self):
shape_format = [
[0, True, 0.44, 0.22, torch.float32],
[1, True, 0.44, 0.22, torch.float32],
[7, True, 0.44, 0.22, torch.float32],
[12, True, 0.44, 0.22, torch.float32],
[7, True, 4.4, 2.2, torch.float32],
[1, True, 4.4, 2.2, torch.float32]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic_alpha_beta(*item)
npu_output = self.npu_op_exec_periodic_alpha_beta(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hamming_window_float16(self):
shape_format = [
[0, torch.float16],
[1, torch.float16],
[7, torch.float16],
[12, torch.float16]]
for item in shape_format:
cpu_output = self.cpu_op_exec(*item)
npu_output = self.npu_op_exec(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hamming_window_periodic_float16(self):
shape_format = [
[0, False, torch.float16],
[1, False, torch.float16],
[7, False, torch.float16],
[12, False, torch.float16]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic(*item)
npu_output = self.npu_op_exec_periodic(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hamming_window_periodic_alpha_float16(self):
shape_format = [
[0, True, 0.22, torch.float16],
[0, True, 2.2, torch.float16],
[1, True, 0.22, torch.float16],
[1, True, 2.0, torch.float16],
[7, True, 0.22, torch.float16],
[7, True, 2.0, torch.float16],
[12, True, 0.22, torch.float16],
[12, True, 2.0, torch.float16],
[0, False, 0.22, torch.float16],
[0, False, 2.2, torch.float16],
[1, False, 2.0, torch.float16],
[7, False, 2.0, torch.float16],
[12, False, 1.1, torch.float16]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic_alpha(*item)
npu_output = self.npu_op_exec_periodic_alpha(*item)
self.assertRtolEqual(cpu_output, npu_output)
def test_hammingwindow_periodic_alpha_beta_float16(self):
shape_format = [
[0, True, 0.44, 0.22, torch.float16],
[1, True, 0.44, 0.22, torch.float16],
[7, True, 0.44, 0.22, torch.float16],
[12, True, 0.44, 0.22, torch.float16],
[7, True, 4.4, 2.2, torch.float16],
[1, True, 4.4, 2.2, torch.float16]]
for item in shape_format:
cpu_output = self.cpu_op_exec_periodic_alpha_beta(*item)
npu_output = self.npu_op_exec_periodic_alpha_beta(*item)
self.assertRtolEqual(cpu_output, npu_output)
if __name__ == "__main__":
run_tests()