fix(//core/conversion/converters/Weights): Fix buffer allocation for weights data #378
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
weights data that occassionally may cause segfaults and causes issues with importing FP16 weights Signed-off-by: Naren Dasan <naren@narendasan.com> Signed-off-by: Naren Dasan <narens@nvidia.com>
github-actions
bot
added
component: conversion
|
@andi4191 can you verify this change fixes the observed segfaults? |
|
I would still say that the compiler should still not be considered thread-safe yet. There's at least a few parts that may still have issues (Logger, Runtime). |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
|
Tested it for a few iterations. It passes everytime. |
andi4191
approved these changes
Feb 25, 2021
| @@ -80,9 +80,29 @@ Weights::Weights(ConversionCtx* ctx, at::Tensor t) { | |||
|
|
|||
| // Store the data in the conversion context so it remains until building is | |||
| // complete | |||
| void* buf = malloc(t_cpu.numel() * sizeof(float)); | |||
|
|
|||
| void* buf; | |||
There was a problem hiding this comment.
Should be initialized to nullptr.
Signed-off-by: Naren Dasan <naren@narendasan.com> Signed-off-by: Naren Dasan <narens@nvidia.com>
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Outdated suggestions cannot be applied.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.
Description
The memcpy for weights tensors over copies when the Tensor is not a FP32 tensor. This seems to be the root cause of segfaults observed in #326 as well as failing DLA tests on aarch64 when users try to compile a model that is already in FP16.
Fixes #326
Type of change
Please delete options that are not relevant and/or add your own.
Checklist: