Added memory optimization for onnx transforms #538
+1,467
−29
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
quic-rishinr
requested review from
ochougul,
quic-hemagnih and
quic-amitraj
as code owners
quic-hemagnih
requested changes
Aug 19, 2025
| from typing import Optional, Tuple | ||
|
|
||
| import numpy as np | ||
| from onnx import ModelProto, external_data_helper, numpy_helper | ||
|
|
||
| from QEfficient.utils.constants import ONNX_TRANSFROM_MEMORY_CLEANUP_INTERVAL | ||
|
|
There was a problem hiding this comment.
NIT: TRANSFORM - spell check
| :param model: The ONNX model to check | ||
| :returns: True if external data is already loaded, False otherwise | ||
| """ | ||
| for tensor in external_data_helper._get_all_tensors(model): |
There was a problem hiding this comment.
Can we think of skipping this extra loop for checking whether for all the tensors external data has been loaded or not. The place where we are loading the external data there we can maintain a flag. This flag by default will be set to false and then once all the external data is loaded we can mark it to TRUE. Then in code we may have to just check the flag. or may not need this function if you want to directly use the flag.
| @@ -61,6 +89,15 @@ def apply(cls, model: ModelProto, *, onnx_base_dir: Optional[str] = None, **kwar | |||
| tensor.CopyFrom(new_tensor) | |||
| transformed = True | |||
|
|
|||
| del neg_inf_mask, clipped_tensor, new_tensor | |||
|
|
|||
There was a problem hiding this comment.
In this loop itself you can check and then update flag
Signed-off-by: Rishin Raj <[email protected]>
Signed-off-by: Rishin Raj <[email protected]>
quic-rishinr
force-pushed
the
export_optim
branch
from
570aad1 to
85c8a0d
Compare
Signed-off-by: Rishin Raj <[email protected]>
Signed-off-by: Rishin Raj <[email protected]>
Signed-off-by: Rishin Raj <[email protected]>
Signed-off-by: Rishin Raj <[email protected]>
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.
Added periodic memory cleanup to FP16ClipTransform and SplitTensorsTransform to reduce memory usage during large tensor processing. Also avoids redundant external data loading when already present.