Add proper support for Binarized and binary neural nets in the QKeras converter #141
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
added 29 commits
March 22, 2024 15:23
…rn new rounding modes.
Fixed rounding_mode specifier in convert_quantized_bits
Commented out assertion on non-representability
…ay_and_auto_po2_scale Merging changes from rounding_mode_new.
…nto HEAD Merge changes from the new converter
…ization_qkeras_converter
…ed up by the normal converter flow. Added an extra InferShape transform
Added a pure binarized networka s a test for qkeras converter
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Outdated suggestions cannot be applied.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 PR cleans up some of the QKeras2QONNX converter code and inserts BinaryQuant nodes instead of Quant nodes for
qkeras.binaryinstances. This PR should follow PR: #139.As this changes the nodes created by the converter it could affect downstream users.
Questions:
channel_order_rewriters._to_channel_first_handler(ctx, quant_act_node). I removed this and the tests still passed, however I am not sure what that was doing there. Does anyone know?# BUG: 1-bit Quant ops currently not exported correctly, manually convert to bipolar valuesSo, should this PR request delete this workaround? or does anyone still depend on this?