[Quantizer] Unsigned Integer Tensor Support #2948
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
djeong20
requested review from
DonghakPark,
EunjuYang,
SeoHyungjun,
again4you,
anyj0527,
baek2sm,
dkjung,
gichan-jang,
haehun,
jaeyun-jung,
jihochu,
jijoongmoon,
leemgs,
lhs8928,
myungjoo,
skykongkong8,
songgot and
wooksong
as code owners
EunjuYang
changed the title
djeong20
force-pushed
the
quantizer/uint/per_tensor
branch
from
997d0b8 to
df61234
Compare
djeong20
changed the title
SeoHyungjun
reviewed
Mar 4, 2025
EunjuYang
reviewed
Mar 4, 2025
nntrainer/tensor/quantizer.cpp
Outdated
Comment on lines
124
to
135
| val = clip(std::lround(input.getValue(b, c, h, w) / *scales), | ||
| quant_min, quant_max); | ||
|
|
||
| if (output.getDataType() == Tdatatype::UINT8 || | ||
| output.getDataType() == Tdatatype::UINT16) { | ||
| val += *zero_points; | ||
| } | ||
|
|
There was a problem hiding this comment.
Is it okay to add zero_point value after clipping?
I find the clip should be applied after adding the zero_point to ensure its range.
There was a problem hiding this comment.
Thank you for the review. min max values are the same as signed integers.
nntrainer/nntrainer/tensor/quantizer.cpp
Lines 43 to 44 in 41fd25b
Therefore, applying zero points after the clipping would produce the correct result.
There was a problem hiding this comment.
I modified it to add a zero point before clipping as we discussed offline. PTAL
There was a problem hiding this comment.
Thank you for the update! It looks more clear and safe! 👍
djeong20
force-pushed
the
quantizer/uint/per_tensor
branch
from
df61234 to
0c23b7c
Compare
EunjuYang
approved these changes
Mar 5, 2025
|
@djeong20 Could you rebase? |
djeong20
force-pushed
the
quantizer/uint/per_tensor
branch
from
0c23b7c to
71957c7
Compare
This pull request enhances the quantizer class by allowing it to perform quantization on tensors with an unsigned integer data type. It includes consideration of the zero-point during the quantization process. **Changes proposed in this PR:** - The quantizer can now quantize and dequantize unsigned integer data types. - Test cases have been added to verify quantization on unsigned integers. - Functions for copying unsigned integer data to floating-point format have also been included. **Self-evaluation:** 1. Build test: [X]Passed [ ]Failed [ ]Skipped 2. Run test: [X]Passed [ ]Failed [ ]Skipped Signed-off-by: Donghyeon Jeong <[email protected]>
djeong20
force-pushed
the
quantizer/uint/per_tensor
branch
from
71957c7 to
c4e7ad5
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
8 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 pull request enhances the quantizer class by allowing it to perform quantization on tensors with an unsigned integer data type. It includes consideration of the zero-point during the quantization process.
Changes proposed in this PR:
Self-evaluation: