github/tinygrad
1
1
Fork 0
mirror of https://github.com/tinygrad/tinygrad.git synced 2026-01-08 22:48:25 -05:00
Code Issues Packages Projects Releases Wiki Activity

switch backward to use gradient [pr] (#8235)

Browse Source 
* switch backward to use gradient [pr]

* set device correctly, dedup

* why does that fail?

* add noop cast

* simple backward

* fix beautiful_mnist

* touchups

* set in compute_gradient

* uop_count

* uop_count was wrong

* collections

* no note

* skip that test

* update sched kernel counts

* train mnist is 65

* fix metadata and gc

* fixes

* materialize_grads

* no pathlib stuff

* add contiguous_backward, fix bugs

* add some realize

* fix multi
This commit is contained in:
George Hotz
2025-01-26 09:12:16 +09:00
committed by GitHub
parent 0ffd572e1e
commit b4bf6a7dea
11 changed files with 50 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
test/test_tensor.py
View File

@@ -770,8 +770,8 @@ class TestTensorMetadata(unittest.TestCase):
self.assertEqual(set(m.name for m in si.metadata), {"relu", "sigmoid", "__mul__"})
def test_complex_backward(self):
x = Tensor.rand(3, requires_grad=True)
y = Tensor.rand(3, requires_grad=True)
x = Tensor.rand(3, requires_grad=True).realize()
y = Tensor.rand(3, requires_grad=True).realize()
out = (x.relu() * y.sigmoid()).sum()
self.assertEqual(out.lazydata.metadata.name, "sum")
out.backward()