This is an automated email from the ASF dual-hosted git repository.
tlopex pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git
The following commit(s) were added to refs/heads/main by this push:
new 5138efcfff [Relax][PyTorch] Unify dtype used in conv2d tests (#18553)
5138efcfff is described below
commit 5138efcfff3585400aaa9566765d67c2c31eb2d8
Author: Guan-Ming (Wesley) Chiu <[email protected]>
AuthorDate: Sat Dec 6 14:16:33 2025 +0800
[Relax][PyTorch] Unify dtype used in conv2d tests (#18553)
## Why
- resolve todo in
[test_op_gradient_numeric.py](https://github.com/apache/tvm/compare/main...guan404ming:update-conv2d-test?expand=1#diff-65bec2fe9ca46b486e6e1d3412e9092d25d3815bb6173435501bbfab7eefd87b)
by unifying the dtype used in conv2d related test
- use float32 with reduced range [0, 3] to maintain numerical precision
for gradient checking
---
tests/python/relax/test_op_gradient_numeric.py | 11 ++++-------
1 file changed, 4 insertions(+), 7 deletions(-)
diff --git a/tests/python/relax/test_op_gradient_numeric.py
b/tests/python/relax/test_op_gradient_numeric.py
index bcea74a883..c76c150f6a 100644
--- a/tests/python/relax/test_op_gradient_numeric.py
+++ b/tests/python/relax/test_op_gradient_numeric.py
@@ -781,11 +781,8 @@ def test_nll_loss_no_batch(target, dev, nll_reduction1,
nll_weighted1, nll_ignor
@tvm.testing.parametrize_targets("llvm")
def test_conv2d(target, dev, c2d_shape1, c2d_shape2, c2d_kwargs):
- # TODO(mlc-team) Update to uniform
- # We should use float32 to check the correctness of conv2d
- # to avoid possible precision problems
- data1_numpy = np.random.uniform(0, 16, c2d_shape1).astype(np.float64)
- data2_numpy = np.random.uniform(0, 3, c2d_shape2).astype(np.float64)
+ data1_numpy = np.random.uniform(0, 3, c2d_shape1).astype(np.float32)
+ data2_numpy = np.random.uniform(0, 3, c2d_shape2).astype(np.float32)
relax_check_gradients(
relax.op.nn.conv2d,
[data1_numpy, data2_numpy],
@@ -819,7 +816,7 @@ def test_conv2d(target, dev, c2d_shape1, c2d_shape2,
c2d_kwargs):
@tvm.testing.parametrize_targets("llvm")
def test_max_pool2d(target, dev, pool_size, pool_kwargs):
- data_numpy = np.random.uniform(0, 16, size=(3, 2, 10,
10)).astype(np.float64)
+ data_numpy = np.random.uniform(0, 3, size=(3, 2, 10,
10)).astype(np.float32)
relax_check_gradients(
relax.op.nn.max_pool2d,
[data_numpy],
@@ -832,7 +829,7 @@ def test_max_pool2d(target, dev, pool_size, pool_kwargs):
@tvm.testing.parametrize_targets("llvm")
def test_avg_pool2d(target, dev, pool_size, pool_kwargs):
- data_numpy = np.random.uniform(0, 16, size=(3, 2, 10,
10)).astype(np.float64)
+ data_numpy = np.random.uniform(0, 3, size=(3, 2, 10,
10)).astype(np.float32)
relax_check_gradients(
relax.op.nn.avg_pool2d,
[data_numpy],
