Skip to content

[Maintain][Manifest] flip Var/Std/VarMean fwd ops to status: implemented #1430

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lcy-seso merged 1 commit into from
May 14, 2026
+3 −3
Merged

[Maintain][Manifest] flip Var/Std/VarMean fwd ops to status: implemented #1430

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions tileops/manifest/reduction.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -404,7 +404,7 @@ ProdFwdOp:
VarFwdOp:
ref_api: "torch.var"
family: reduction
status: spec-only # dim=None not yet implemented; update when impl lands
status: implemented
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot May 12, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

While flipping the status to implemented is correct, I noticed a bug in the underlying implementation that will affect these ops. In tileops/ops/reduction/reduce.py, the eval_roofline method (line 180) attempts to access self.dtype.itemsize. However, torch.dtype objects do not have an itemsize attribute; the correct method is element_size(). This will cause an AttributeError when calculating roofline metrics for these newly implemented ops.

Copy link
Copy Markdown
Collaborator Author

@lcy-seso lcy-seso May 12, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Verified false positive. torch.dtype.itemsize has existed since torch 2.1, and pyproject.toml pins torch>=2.1.0,<2.11.0. Local check on torch 2.9.1: torch.float16.itemsize == 2, torch.int64.itemsize == 8 — no AttributeError. No fix needed.

Comment thread
lcy-seso marked this conversation as resolved.
Comment thread
lcy-seso marked this conversation as resolved.
Comment thread
lcy-seso marked this conversation as resolved.

signature:
inputs:
Expand Down Expand Up @@ -449,7 +449,7 @@ VarFwdOp:
StdFwdOp:
ref_api: "torch.std"
family: reduction
status: spec-only # dim=None not yet implemented; update when impl lands
status: implemented
Comment thread
lcy-seso marked this conversation as resolved.
Comment thread
lcy-seso marked this conversation as resolved.

signature:
inputs:
Expand Down Expand Up @@ -494,7 +494,7 @@ StdFwdOp:
VarMeanFwdOp:
ref_api: "torch.var_mean"
family: reduction
status: spec-only # dim=None not yet implemented; update when impl lands
status: implemented
Comment thread
lcy-seso marked this conversation as resolved.
Comment thread
lcy-seso marked this conversation as resolved.

signature:
inputs:
Expand Down
Loading