"""
"""
import os
import math
import pypto
import torch
def test_remainder_onboard():
device_id = int(os.environ.get('TILE_FWK_DEVICE_ID', 0))
torch.npu.set_device(device_id)
shape = (72, 71)
view_shape = (32, 32)
tile_shape = (32, 32)
pypto.runtime._device_init()
input1 = pypto.tensor(shape, pypto.DT_INT16, "PTO_TENSOR_input1")
input2 = pypto.tensor(shape, pypto.DT_INT16, "PTO_TENSOR_input2")
output = pypto.tensor(shape, pypto.DT_INT16, "PTO_TENSOR_output")
b_loop_num = math.ceil(shape[0] / view_shape[0])
s_loop_num = math.ceil(shape[1] / view_shape[1])
with pypto.function("MAIN", input1, input2, output):
for b_idx in pypto.loop(b_loop_num, name="b0", idx_name="bidx"):
for s_idx in pypto.loop(s_loop_num, name="s0", idx_name="sidx"):
valid_shape = [pypto.min(pypto.symbolic_scalar(shape[0]) - b_idx * view_shape[0],
pypto.symbolic_scalar(view_shape[0])),
pypto.min(pypto.symbolic_scalar(shape[1]) - s_idx * view_shape[1],
pypto.symbolic_scalar(view_shape[1]))]
offsets = [b_idx * view_shape[0], s_idx * view_shape[1]]
view_tensor_a = pypto.view(input1, view_shape, offsets, valid_shape=valid_shape)
view_tensor_b = pypto.view(input2, view_shape, offsets, valid_shape=valid_shape)
pypto.set_vec_tile_shapes(tile_shape[0], tile_shape[1])
view_tensor_c = pypto.remainder(view_tensor_a, view_tensor_b)
pypto.assemble(view_tensor_c, offsets, output)
del view_tensor_a, view_tensor_b, view_tensor_c
a_tensor = torch.randint(
low=1, high=100, size=[shape[0], shape[1]], dtype=torch.int16)
b_tensor = torch.randint(
low=-100, high=-1, size=[shape[0], shape[1]], dtype=torch.int16)
out_tensor = torch.zeros(shape[0], shape[1], dtype=torch.int16)
pto_a_tensor = pypto.from_torch(a_tensor, "a_tensor")
pto_b_tensor = pypto.from_torch(b_tensor, "b_tensor")
pto_out_tensor = pypto.from_torch(out_tensor, "out_tensor")
pypto.runtime._device_run_once_data_from_host(pto_a_tensor, pto_b_tensor, pto_out_tensor)
golden = torch.remainder(a_tensor, b_tensor)
assert torch.allclose(out_tensor.flatten(), golden.flatten(), rtol=1e-4, atol=1e-5)
pypto.runtime._device_fini()
def test_remainders_onboard():
device_id = int(os.environ.get('TILE_FWK_DEVICE_ID', 0))
torch.npu.set_device(device_id)
shape = (72, 71)
scalar = 3
view_shape = (32, 32)
tile_shape = (16, 16)
pypto.runtime._device_init()
input1 = pypto.tensor(shape, pypto.DT_INT16, "PTO_TENSOR_input1")
output = pypto.tensor(shape, pypto.DT_INT16, "PTO_TENSOR_output")
b_loop_num = math.ceil(shape[0] / view_shape[0])
s_loop_num = math.ceil(shape[1] / view_shape[1])
with pypto.function("MAIN", input1, output):
for b_idx in pypto.loop(b_loop_num, name="b0", idx_name="bidx"):
for s_idx in pypto.loop(s_loop_num, name="s0", idx_name="sidx"):
valid_shape = [pypto.min(pypto.symbolic_scalar(shape[0]) - b_idx * view_shape[0],
pypto.symbolic_scalar(view_shape[0])),
pypto.min(pypto.symbolic_scalar(shape[1]) - s_idx * view_shape[1],
pypto.symbolic_scalar(view_shape[1]))]
offsets = [b_idx * view_shape[0], s_idx * view_shape[1]]
view_tensor_a = pypto.view(input1, view_shape, offsets, valid_shape=valid_shape)
pypto.set_vec_tile_shapes(tile_shape[0], tile_shape[1])
view_tensor_b = pypto.remainder(view_tensor_a, scalar)
pypto.assemble(view_tensor_b, offsets, output)
del view_tensor_a, view_tensor_b
a_tensor = torch.randint(
low=1, high=100, size=[shape[0], shape[1]], dtype=torch.int16)
out_tensor = torch.zeros(shape[0], shape[1], dtype=torch.int16)
pto_a_tensor = pypto.from_torch(a_tensor, "a_tensor")
pto_out_tensor = pypto.from_torch(out_tensor, "out_tensor")
pypto.runtime._device_run_once_data_from_host(pto_a_tensor, pto_out_tensor)
golden = torch.remainder(a_tensor, scalar)
assert torch.allclose(out_tensor.flatten(), golden.flatten(), rtol=1e-4, atol=1e-5)
pypto.runtime._device_fini()
def test_remainderrs_onboard():
device_id = int(os.environ.get('TILE_FWK_DEVICE_ID', 0))
torch.npu.set_device(device_id)
shape = (72, 71)
scalar = 3.2
view_shape = (32, 32)
tile_shape = (16, 16)
pypto.runtime._device_init()
input1 = pypto.tensor(shape, pypto.DT_FP32, "PTO_TENSOR_input1")
output = pypto.tensor(shape, pypto.DT_FP32, "PTO_TENSOR_output")
b_loop_num = math.ceil(shape[0] / view_shape[0])
s_loop_num = math.ceil(shape[1] / view_shape[1])
with pypto.function("MAIN", input1, output):
for b_idx in pypto.loop(b_loop_num, name="b0", idx_name="bidx"):
for s_idx in pypto.loop(s_loop_num, name="s0", idx_name="sidx"):
valid_shape = [pypto.min(pypto.symbolic_scalar(shape[0]) - b_idx * view_shape[0],
pypto.symbolic_scalar(view_shape[0])),
pypto.min(pypto.symbolic_scalar(shape[1]) - s_idx * view_shape[1],
pypto.symbolic_scalar(view_shape[1]))]
offsets = [b_idx * view_shape[0], s_idx * view_shape[1]]
view_tensor_a = pypto.view(input1, view_shape, offsets, valid_shape=valid_shape)
pypto.set_vec_tile_shapes(tile_shape[0], tile_shape[1])
view_tensor_b = pypto.remainder(scalar, view_tensor_a)
pypto.assemble(view_tensor_b, offsets, output)
del view_tensor_a, view_tensor_b
a_tensor = torch.randn(shape, dtype=torch.float32)
out_tensor = torch.zeros(shape[0], shape[1], dtype=torch.float32)
pto_a_tensor = pypto.from_torch(a_tensor, "a_tensor")
pto_out_tensor = pypto.from_torch(out_tensor, "out_tensor")
pypto.runtime._device_run_once_data_from_host(pto_a_tensor, pto_out_tensor)
golden = torch.remainder(scalar, a_tensor)
assert torch.allclose(out_tensor.flatten(), golden.flatten(), rtol=1e-4, atol=1e-5)
pypto.runtime._device_fini()