Repository navigation
[Bug][Relax] Inconsistent IRs will be got when Load IRs with R.nn.pad operator #17483
Description
Actual behavior
error: pad() got multiple values for argument 'pad_width'
--> <str>:8:55
|
8 | gv: R.Tensor((2, 130, 30), dtype="float32") = R.nn.pad(x, R.const(0, "int32"), pad_width=[0, 0, 1, 1, 1, 1], pad_mode="constant")
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Steps to reproduce
import tvm
from tvm import relax
from tvm.script import relax as R
@tvm.script.ir_module
class Module:
@R.function
def main(x: R.Tensor((2, 128, 28), "float32")) -> R.Tensor((2, 130, 30), "float32"):
gv: R.Tensor((2, 130, 30), "float32") = R.nn.pad(x, (0, 0, 1, 1, 1, 1))
return gv
m = Module
m.show()
m_str = m.script()
m2 = tvm.script.from_source(m_str)
m2.show() # crash
Activity
[-][Bug][Relax] The `script()` API output wrong IRs with R.nn.pad operator[/-][+][Bug][Relax] Inconsistent IRs will be got when Load IRs with R.nn.pad operator[/+]yongwww commentedon Oct 24, 2024
looks the nn.pad is not roundtrip-able with tvmscript, the ordering the parameters in parser is different from that in printer.
jikechao commentedon Oct 26, 2024
@yongwww Agree! Should we provide a PR to fix the
nn.padto a roundtrip-able operator?BTW,
R.nn.attentionhas a similar issue: #17486yongwww commentedon Oct 26, 2024
@jikechao I'd appreciate it if you could submit a PR to fix this. No worries if you're busy, please let me know and I'll take care of it.
jikechao commentedon Oct 27, 2024
@yongwww I am not very familiar with the relevant source code, it would be better to assign it to other colleagues. Thanks!
yongwww commentedon Feb 17, 2025
@jikechao the issue should be fixed now after #17657 was landed, thanks for raising the issue, and thanks @parsifal-47 for the great patch!