[Frontend] Prevent tflite frontend from producing int64 shape/parameters#7030
Merged
tqchen merged 2 commits intoapache:mainfrom Dec 4, 2020
Merged
[Frontend] Prevent tflite frontend from producing int64 shape/parameters#7030tqchen merged 2 commits intoapache:mainfrom
tqchen merged 2 commits intoapache:mainfrom
Conversation
merrymercy
changed the title
Member
|
Seems this pr should resolve #7009 goal |
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Dec 4, 2020
trevor-m
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Dec 4, 2020
electriclilies
pushed a commit
to electriclilies/tvm
that referenced
this pull request
Feb 18, 2021
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
I found some shapes and parameters (padding) are converted to int64 from tflite frontend.
The consequence is that the shape of a placeholder becomes
[1, (int64) 112, (int64) 112, 64]. Some of them are int64, the others are int32.This type-mixed shape fails some optimization in the auto-scheduler.
The reason for int64 shape is
ShapeAsNumpyreturnsnp.int32, and it is automatically upcasted tonp.int64when doing computation. To fix this, I convert allnp.int32to python's int to prevent auto upcast.