forked from tinygrad/tinygrad
120 lines
3.7 KiB
Python
120 lines
3.7 KiB
Python
#!/usr/bin/env python
|
|
import gc, inspect
|
|
import unittest
|
|
import numpy as np
|
|
from tinygrad.device import Buffer
|
|
from tinygrad.engine.realize import run_linear
|
|
from tinygrad.uop.ops import UOp
|
|
from tinygrad.tensor import Tensor
|
|
|
|
def _allocations_of_type(t):
|
|
ret = 0
|
|
for x in gc.get_objects():
|
|
try:
|
|
if isinstance(x, t): ret += 1
|
|
except ReferenceError:
|
|
pass
|
|
return ret
|
|
|
|
def tensors_allocated():
|
|
gc.collect()
|
|
return _allocations_of_type(Tensor)
|
|
|
|
def bufs_allocated():
|
|
# count Buffer objects that own storage: a realized (or to-be-realized) BUFFER UOp owns one, views are transient and excluded
|
|
gc.collect()
|
|
return sum(1 for x in gc.get_objects() if isinstance(x, Buffer) and x._base is None)
|
|
|
|
class TestGC(unittest.TestCase):
|
|
|
|
def test_gc(self):
|
|
Tensor.manual_seed(0)
|
|
base = tensors_allocated()
|
|
a = Tensor.rand(4, 4)
|
|
b = Tensor.zeros(4, 4)
|
|
(a*b).mean().backward()
|
|
assert (tensors_allocated()-base > 0)
|
|
del a,b
|
|
assert (tensors_allocated()-base == 2) # one for Tensor._device_rng_counters, and one for Tensor._device_seeds
|
|
Tensor.manual_seed(0)
|
|
|
|
def test_gc_complex(self):
|
|
Tensor.manual_seed(0)
|
|
base = tensors_allocated()
|
|
a = Tensor(np.zeros((4, 4), dtype=np.float32))
|
|
b = Tensor.rand(4, 4)
|
|
assert (tensors_allocated()-base == 4)
|
|
(a*b).mean().backward()
|
|
assert (tensors_allocated()-base == 6)
|
|
del b
|
|
assert (tensors_allocated()-base == 4)
|
|
b = Tensor(np.zeros((4, 4), dtype=np.float32))
|
|
print(tensors_allocated())
|
|
(a*b).mean().backward()
|
|
print(tensors_allocated())
|
|
assert (tensors_allocated()-base == 6)
|
|
del b
|
|
assert (tensors_allocated()-base == 4)
|
|
Tensor.manual_seed(0)
|
|
|
|
def test_schedule_gc(self):
|
|
init = bufs_allocated()
|
|
x = Tensor.ones(256).contiguous().realize()
|
|
y = Tensor.ones(5, 5).contiguous()
|
|
y.schedule_linear()
|
|
del x
|
|
del y
|
|
self.assertEqual(bufs_allocated()-init, 0)
|
|
|
|
def test_schedule_gc_with_inputs(self):
|
|
init = bufs_allocated()
|
|
x = Tensor.ones(256).contiguous().realize()
|
|
y = x+Tensor.ones(256).contiguous()
|
|
del x
|
|
run_linear(*y.linear_with_vars())
|
|
self.assertEqual(bufs_allocated()-init, 1)
|
|
del y
|
|
self.assertEqual(bufs_allocated()-init, 0)
|
|
|
|
def test_toposort_blocks_gc(self):
|
|
init = bufs_allocated()
|
|
x = Tensor.ones(4,4).contiguous().realize()+1
|
|
self.assertEqual(bufs_allocated()-init, 1)
|
|
# try commenting this part out, it's green!
|
|
x.uop.toposort()
|
|
del x
|
|
if bufs_allocated()-init != 0:
|
|
print(inspect.getclosurevars(UOp.toposort().fget))
|
|
raise AssertionError(f"never gced {[x for x in gc.get_objects() if isinstance(x, Buffer)]}")
|
|
|
|
def test_buffer_ownership(self):
|
|
init = bufs_allocated()
|
|
a = Tensor.empty(10)
|
|
# the Buffer object is owned by the BUFFER UOp 1:1, it exists from creation (device memory is still allocated lazily)
|
|
self.assertEqual(bufs_allocated()-init, 1)
|
|
a.realize()
|
|
real_buf = a.uop.buffer
|
|
self.assertIs(a.uop.arg.buffer, real_buf)
|
|
self.assertEqual(bufs_allocated()-init, 1)
|
|
del a.uop
|
|
self.assertEqual(bufs_allocated()-init, 1) # the Buffer object is still held here
|
|
del real_buf
|
|
self.assertEqual(bufs_allocated()-init, 0)
|
|
|
|
def test_assign_keeps_buffer(self):
|
|
init = bufs_allocated()
|
|
a = Tensor.full((4,), 1.).contiguous()
|
|
a.realize()
|
|
real_buf = a.uop.buffer
|
|
a.assign(Tensor.full((4,), 2.))
|
|
# assign writes in place: the AFTER still references the same Buffer
|
|
self.assertIs(a.uop.src[0].buffer, real_buf)
|
|
a.realize()
|
|
del a
|
|
self.assertEqual(bufs_allocated()-init, 1) # the Buffer object is still held here
|
|
del real_buf
|
|
self.assertEqual(bufs_allocated()-init, 0)
|
|
|
|
if __name__ == '__main__':
|
|
unittest.main()
|