mshr-h commented on code in PR #17841:
URL: https://github.com/apache/tvm/pull/17841#discussion_r2048142477
##########
tests/python/relax/test_frontend_from_fx.py:
##########
@@ -4513,5 +4513,134 @@ def main(
verify_model(Narrow(), [([5, 3], "float32")], {}, Expected)
+def test_norm():
+
+ input_info = [([1, 3, 5, 3], "float32")]
+
+ class Norm(Module):
+ def __init__(self, p, dim=None, keepdim=False):
+ super().__init__()
+ self.p = p
+ self.dim = dim
+ self.keepdim = keepdim
+
+ def forward(self, x):
+ return torch.norm(x, p=self.p, dim=self.dim, keepdim=self.keepdim)
+
+ @tvm.script.ir_module
+ class Expected1:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((), dtype="float32") = R.max(R.abs(inp_0),
axis=None, keepdims=False)
+ gv: R.Tensor((), dtype="float32") = lv
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected2:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((), dtype="float32") = R.min(R.abs(inp_0),
axis=None, keepdims=False)
+ gv: R.Tensor((), dtype="float32") = lv
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected3:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((1, 3, 5, 3), dtype="float32") = R.abs(inp_0)
+ lv1: R.Tensor((1, 3, 5, 3), dtype="float32") = R.power(lv,
R.const(2, "float32"))
+ lv2: R.Tensor((), dtype="float32") = R.sum(lv1, axis=None,
keepdims=False)
+ lv3: R.Tensor((), dtype="float32") = R.power(lv2, R.const(0.5,
"float32"))
+ gv: R.Tensor((), dtype="float32") = lv3
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected4:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((1, 3, 5, 3), dtype="float32") = R.abs(inp_0)
+ lv1: R.Tensor((1, 3, 5, 3), dtype="float32") = R.power(lv,
R.const(1.0, "float32"))
+ lv2: R.Tensor((), dtype="float32") = R.sum(lv1, axis=None,
keepdims=False)
+ lv3: R.Tensor((), dtype="float32") = R.power(lv2, R.const(1.0,
"float32"))
+ gv: R.Tensor((), dtype="float32") = lv3
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected5:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((1, 3, 5, 3), dtype="float32") = R.abs(inp_0)
+ lv1: R.Tensor((1, 3, 5, 3), dtype="float32") = R.power(lv,
R.const(-4, "float32"))
+ lv2: R.Tensor((), dtype="float32") = R.sum(lv1, axis=None,
keepdims=False)
+ lv3: R.Tensor((), dtype="float32") = R.power(lv2,
R.const(-0.25, "float32"))
+ gv: R.Tensor((), dtype="float32") = lv3
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected6:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((1, 3, 5, 3), dtype="float32") = R.abs(inp_0)
+ lv1: R.Tensor((1, 3, 5, 3), dtype="float32") = R.power(lv,
R.const(0.5, "float32"))
+ lv2: R.Tensor((), dtype="float32") = R.sum(lv1, axis=None,
keepdims=False)
+ lv3: R.Tensor((), dtype="float32") = R.power(lv2, R.const(2,
"float32"))
+ gv: R.Tensor((), dtype="float32") = lv3
+ R.output(gv)
+ return gv
+
+ @tvm.script.ir_module
+ class Expected7:
+ @R.function
+ def main(
+ inp_0: R.Tensor((1, 3, 5, 3), dtype="float32"),
+ ) -> R.Tensor((), dtype="float32"):
+ with R.dataflow():
+ lv: R.Tensor((1, 3, 5, 3), dtype="float32") =
R.multiply(inp_0, inp_0)
+ lv1: R.Tensor((), dtype="float32") = R.sum(lv, axis=None,
keepdims=False)
+ lv2: R.Tensor((), dtype="float32") = R.sqrt(lv1)
+ gv: R.Tensor((), dtype="float32") = lv2
+ R.output(gv)
+ return gv
+
+ norms = [
+ (float("inf"), None, False),
Review Comment:
just for your info, you can define something like this for test cases.
```python
norms = [
((float("inf"), None, False), Expected1),
'(float("-inf"), None, False), Expected2),
```
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
