SYCL: set extras only on GGML_TYPE_Q4_0 #12366
Merged
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
ggml
Member
|
The extras should also be freed in the |
qnixsynapse
force-pushed
the
fix/memory_leak
branch
from
fb5fefd to
0cee320
Compare
Collaborator
Author
|
@NeoZhangJianyu Can you review this PR please? |
NeoZhangJianyu
approved these changes
Mar 17, 2025
Collaborator
NeoZhangJianyu
left a comment
NeoZhangJianyu
left a comment
There was a problem hiding this comment.
It's great!
Thank you!
arthw
pushed a commit
to arthw/llama.cpp
that referenced
this pull request
Mar 19, 2025
* SYCL: set extras only on GGML_TYPE_Q4_0 * release tensor_extras in reset buffer interface
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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Outdated suggestions cannot be applied.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.
Commit 08d5986 implemented optimization of Q4_0 tensors on intel GPUs by opting to reorder the Q4 block to separate quantized weights and dequantize scaler.
However, since this commit required setting extras in init_tensor function, this commit did not check if the tensor type is indeed of Q4_0, which resulted in memory leak.
This change adds a condition to prevent memory leak.
ps. This is not a permanent solution. We should remove setting extras inside init_tensor function.
Tested with both non Q4_0 and Q4_0 models.