Skip to content

Missing kTfLiteInt16 in ComparisonsPrepare #3391

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
veblush wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

Missing kTfLiteInt16 in ComparisonsPrepare #3391

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion tensorflow/lite/micro/kernels/comparisons.cc
View file
Open in desktop
Copy link
Member

@ddavis-2015 ddavis-2015 Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@veblush The actual comparison Eval methods should also have INT16 support added. Not all LiteRT comparisons have INT16 support, and it is confusing how some LiteRT comparisons assume (or don't) that the INT16 inputs are not quantized.

Original file line number Diff line number Diff line change
Expand Up @@ -559,7 +559,7 @@ TfLiteStatus ComparisonsPrepare(TfLiteContext* context, TfLiteNode* node) {
micro_context->AllocateTempInputTensor(node, kInputTensor2);
TF_LITE_ENSURE(context, input2 != nullptr);

if (input1->type == kTfLiteInt8) {
if (input1->type == kTfLiteInt8 || input1->type == kTfLiteInt16) {
Copy link
Member

@ddavis-2015 ddavis-2015 Feb 8, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@veblush should also have a check if the tensor is in fact quantized (unless the TfLite converter always produces quantized INT16???). I always assumed un-quantized INT16 was just as valid as unquantized INT32. Also add an is_quantized bool to the OpData?

auto input1_offset = -input1->params.zero_point;
auto input2_offset = -input2->params.zero_point;
const int kLeftShift = 8;
Expand Down
Loading