[Relax][PyTorch] Fix AssertionError: Unsupported function types randn.default
#18396
+32
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
cchung100m
force-pushed
the
issue-18362
branch
8 times, most recently
from
ed0fb06 to
425e78f
Compare
cchung100m
force-pushed
the
issue-18362
branch
from
425e78f to
6404d3b
Compare
cchung100m
force-pushed
the
issue-18362
branch
from
fef7b14 to
2ff97f7
Compare
Contributor
Author
tlopex
reviewed
Jan 4, 2026
| dtype = self._convert_data_type( | ||
| node.kwargs.get("dtype", torch.get_default_dtype()), self.env | ||
| ) | ||
| return self.block_builder.emit(relax.op.zeros(node.args[0], dtype)) |
Member
There was a problem hiding this comment.
Could you add a TODO note here indicates that we still need to implement it properly?
Contributor
|
I don't know if we really need a temp support for rand. As we added custom op converter in #18544 , users can now define randn conveter in their codebase instead of editing the tvm torch frontend files. |
Contributor
Author
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR is trying to fix issues #18362.
Solution:
For now, return zeros with the correct shape since random initialization randn is mainly used during training, not inference. This is a temporary solution until Relax adds proper random number generation support.
Future Enhancement:
A proper implementation would require adding:
relax.op.random_normal(shape, mean=0, std=1, dtype, seed)operator