[Torch] Fold aten rounding ops on splat constants. #4359
+191
−12
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
This commit teaches the folding methods of `AtenFloor`, `AtenCeil`, `AtenRound`, and `AtenTruc` to constant-fold roundings when the operand is a splat `DenseElementsAttr`.
mdazz
force-pushed
the
mdazz/ad-rounding-folder
branch
from
58550c3 to
175d2a2
Compare
|
@sahas3 @zjgarvey @vivekkhandelwal1 can you please take a look at this? |
sahas3
reviewed
Nov 5, 2025
| } | ||
|
|
||
| // Common helper for splat-only rounding-based folders. | ||
| static OpFoldResult foldSplatRounding(ValueTensorType resultType, |
There was a problem hiding this comment.
I think naming it foldFloatSplatWithRounding will be more appropriate since it only handles float data.
| return {}; | ||
|
|
||
| auto outShaped = resultType.toBuiltinTensor(); | ||
| if (!outShaped.hasStaticShape()) |
There was a problem hiding this comment.
why is static shape a requirement?
| @export | ||
| @annotate_args([None]) | ||
| def forward(self): | ||
| return torch.ops.aten.round(self.const) |
There was a problem hiding this comment.
I am guessing the other ops are already covered in e2e tests?
| // CHECK: %[[C:.*]] = torch.vtensor.literal(dense<-1.000000e+00> : tensor<2x2xf32>) | ||
| // CHECK: return %[[C]] | ||
| func.func @torch.aten.ceil$fold() -> !torch.vtensor<[2,2],f32> { | ||
| %cst = torch.vtensor.literal(dense<-1.100000e+00> : tensor<2x2xf32>) |
There was a problem hiding this comment.
Can you add a +ve value too for completeness?
| // CHECK: %[[C:.*]] = torch.vtensor.literal(dense<1.000000e+00> : tensor<3x4xf32>) | ||
| // CHECK: return %[[C]] | ||
| func.func @torch.aten.floor$fold() -> !torch.vtensor<[3,4],f32> { | ||
| %cst = torch.vtensor.literal(dense<1.900000e+00> : tensor<3x4xf32>) |
There was a problem hiding this comment.
Similarly adding a negative value will give full coverage
Comment on lines
+253
to
+254
| // NaNs and infs are dealt with consistently with torch, so side-effects | ||
| // can be discarded. |
There was a problem hiding this comment.
Clarification: Can you elaborate on what you meant by "NaNs and infs are dealt with consistently with torch" ?
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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 commit teaches the folding methods of
AtenFloor,AtenCeil,AtenRound, andAtenTructo constant-fold roundings when the operand is a splatDenseElementsAttr.