Skip to content
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

Jump to bottom

[res] Add TFLiteRecipe for dynamic shape pad #14109

Merged
merged 2 commits into from
Sep 27, 2024
Merged

[res] Add TFLiteRecipe for dynamic shape pad #14109

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
652dfce
[res] Add TFLiteRecipe for dynamic shape pad
icodo98 Sep 27, 2024
253a35a
Update res/TensorFlowLiteRecipes/Inf_Pad_000/test.recipe
icodo98 Sep 27, 2024
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
33 changes: 33 additions & 0 deletions res/TensorFlowLiteRecipes/Inf_Pad_000/test.recipe
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,33 @@
# padding with dynamic shape, others same as Pad_000
operand {
name: "ifm"
type: FLOAT32
shape { dim: 1 dim: 1 dim: 3 dim: 2 }
Copy link
Contributor

Choose a reason for hiding this comment

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

Just curious,
Is there any special reason to set the second dim to 1?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Just curious, Is there any special reason to set the second dim to 1?

There is no particular reason. I just wanted to check if rule-lib.sh works well in dims other than the first one.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Just curious, Is there any special reason to set the second dim to 1?

Oh, I misunderstand your question. 😂

I referred this code for shape_signature to test second dim to dynamic.

ONE/compiler/tflchef/tests/shape_signature/test.recipe

Lines 4 to 5 in 06a536b

shape { dim: 1 dim: 3 dim: 3 dim: 2 }
shape_signature { dim: -1 dim: 3 dim: 3 dim: 2 }

shape_signature { dim: 1 dim: -1 dim: 3 dim: 2 }
}
operand {
name: "padding"
type: INT64
icodo98 marked this conversation as resolved.
Show resolved Hide resolved
shape { dim: 4 dim: 2 }
filler {
tag: "explicit"
arg: "0" arg: "0"
arg: "1" arg: "1"
arg: "2" arg: "2"
arg: "0" arg: "0"
}
}
operand {
name: "ofm"
type: FLOAT32
shape { dim: 1 dim: 1 dim: 7 dim: 2 }
shape_signature { dim: 1 dim: -1 dim: 7 dim: 2 }
}
operation {
type: "Pad"
input: "ifm"
input: "padding"
output: "ofm"
}
input: "ifm"
output: "ofm"
5 changes: 5 additions & 0 deletions res/TensorFlowLiteRecipes/Inf_Pad_000/test.rule
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
# To check if dynamic dimension properly inferred

RULE "VERIFY_FILE_FORMAT" $(verify_file_format) '=' 1

RULE "PAD_SHAPE" $(tensor_shape ofm) '=' [1,-1,7,2]