Lunderberg commented on code in PR #16785:
URL: https://github.com/apache/tvm/pull/16785#discussion_r1539406497
##########
python/tvm/relax/frontend/nn/op.py:
##########
@@ -676,12 +676,31 @@ def permute_dims(x: Tensor, axes: Optional[List[int]] =
None, name: str = None)
result : Tensor
The transposed result.
"""
+
+ # TODO(Lunderberg): This is a more extensive auto-naming than
+ # intended here. Is this still worth it?
Review Comment:
Long-term, I want to move this automatic naming from the `nn.Module` side to
the Relax side, since it could then be performed after removal of trivial
bindings. I don't expect these chains to be deep, as it only tracks trivial
bindings. The trivial binding from the Relax function parameter to the
parameter's `param._expr` field should be the only one that would be tracked.
--
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]
