Skip to content

quantization script updates#2208

Merged
askhade merged 1 commit intomasterfrom
askhade/bias_quantization_calibration
Oct 21, 2019
Merged

quantization script updates#2208
askhade merged 1 commit intomasterfrom
askhade/bias_quantization_calibration

Conversation

@askhade
Copy link
Contributor

@askhade askhade commented Oct 21, 2019 •
edited
Loading

Description: Includes updates to quantization script

  1. Bias quantization for QlinearConv and ConvInteger nodes
  2. Reuse already quantized nodes
  3. Discard activation nodes which follow QlinearConv nodes
  4. Better clean-up after quantization

Motivation and Context

  • Earlier conv nodes with bias could not be quantized correctly. This PR fixes this issue.
  • Earlier quantization was done for every possible node... now there is better reuse of quantized initializers and QuantizeLinear and DequantizeLinear nodes are added only when required.

@askhade askhade requested a review from a team as a code owner October 21, 2019 17:08
@askhade
Copy link
Contributor Author

askhade commented Oct 21, 2019 •
edited
Loading

This PR includes work from PRs submitted to askhade/quantization_and_caliberation branch
#1827
#1794
#1762

@askhade askhade merged commit d197079 into master Oct 21, 2019
@askhade askhade deleted the askhade/bias_quantization_calibration branch October 21, 2019 19:26
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

1 more reviewer

@snnn snnn snnn approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@askhade @snnn