adding rotary embedding example, with graph rewrite for complex subgraph [WIP] #3570
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
github-actions
bot
added
component: lowering
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
There was a problem hiding this comment.
There are some changes that do not conform to Python style guidelines:
--- /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/constant_folding.py 2025-06-13 00:06:13.440339+00:00
+++ /home/runner/work/TensorRT/TensorRT/py/torch_tensorrt/dynamo/lowering/passes/constant_folding.py 2025-06-13 00:06:42.189786+00:00
@@ -98,16 +98,17 @@
class _TorchTensorRTConstantFolder(ConstantFolder): # type: ignore[misc]
def __init__(self, *args: Any, **kwargs: Any) -> None:
super().__init__(*args, **kwargs)
def is_impure(self, node: torch.fx.node.Node) -> bool:
- # Set of known quantization ops to be excluded from constant folding.
+ # Set of known quantization ops to be excluded from constant folding.
# Currently, we exclude all quantization ops coming from modelopt library.
quantization_ops = {}
try:
- # modelopt import ensures torch.ops.tensorrt.quantize_op.default is registered
+ # modelopt import ensures torch.ops.tensorrt.quantize_op.default is registered
import modelopt.torch.quantization as mtq
+
assert torch.ops.tensorrt.quantize_op.default
quantization_ops.add(torch.ops.tensorrt.quantize_op.default)
except Exception as e:
pass
if quantization_ops and node.target in quantization_ops:
peri044
reviewed
Jun 13, 2025
Comment on lines
+247
to
+255
| """Constructs the original and replacement functions for complex multiplication. | ||
|
|
||
| The original functions correspond to native complex multiplication | ||
| via torch.mul or operator.mul on complex tensors. | ||
|
|
||
| The replacement function assumes x and y are real tensors with the last | ||
| dimension size 2 representing real and imaginary parts, and performs | ||
| complex multiplication manually returning the same shaped tensor. | ||
| """ |
There was a problem hiding this comment.
if you know the limitations of this pass, please add for future reference
| class ComplexOpDetector: | ||
| def __init__(self, logger): | ||
| self.logger = logger | ||
| pass |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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-
Please note that this PR is currently for the single GPU case where there is no DTensor in the inputs of the torch module. Ideally this should not require runtime changes.
This should avoid the graph breaks caused due to view_as_complex and view_as_real nodes.