Support u8 and i8 tensors in operator inputs, outputs and model files #345
Conversation
This file contains 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
In preparation for adding new int tensor types, rename the existing ones to specify the size.
This will make adding new input types easier.
Initially these are not supported by any operators. They can only be created and passed around as graph constants and values.
When adding a new tensor, the returned offset was the end of the previous tensor instead of the start of the new tensor. These offsets may differ if padding bytes are added for alignment. The problem went unnoticed before because all the supported data types had the same size (32bit), and that has now changed.
robertknight
force-pushed
the
rten-model-u8i8
branch
from
d8d1518
to
387cdb8
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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 is the first step in adding 8-bit quantization support. It adds support for u8 and i8 tensors in:
Int8 support has not been added to most operators yet. Done naively this would require two instantiations of the code for u8 and i8 tensors even though many operators could use the same code for both. This could be addressed with #244. For the moment I am adding support for i8 and u8 tensors to operators on an as-needed basis.