Use IDs for placeholder/outputs of implicit q/dq nodes #141
Conversation
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
Summary: X-link: pytorch/pytorch#107930 att bypass-github-export-checks Reviewed By: kimishpatel Differential Revision: D48588121 fbshipit-source-id: a6dd25bfd70d2f54202b497fa68cac2ef8b9626a
Differential Revision: D48667679 fbshipit-source-id: c1365045f9a93609d7a07e5229a2799af05a3f47
Summary:
In XNNPACK from the partitioner we encounter quantize node that are outputs to the graph. Since these quantize nodes are implicit and not actually quantizing the values, we must give its externalness to the node with which it quantizes.
### Outputs
```
|
conv --> q ----> | ----> dq -->
|
Delegate boundary
```
In this example the output of the delegate is the output of quantized Conv. In XNNPACK, we must designate the output of Conv to be an external value. In this graph, however, the q node is the output. Since q is an implicit node (it does not actually running quantize) We pass its externalness to the conv node.
### Inputs
i.e.
```
|
q ----> | ----> dq --> conv -->
|
Delegate boundary
```
In this example q is the placeholder node and is given as input to the first dq. But we are running a quantized conv within the delegate. Since dq is implicit, we pass the externalness of the q place holder to dq, so that the input to Conv is external.
Reviewed By: digantdesai
Differential Revision: D48667676
fbshipit-source-id: a6adfdd917a0efd4de56549652900ea614cb9661
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D48667676 |
|
This pull request has been merged in 76b1385. |
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.
Summary:
In XNNPACK from the partitioner we encounter quantize node that are outputs to the graph. Since these quantize nodes are implicit and not actually quantizing the values, we must give its externalness to the node with which it quantizes.
Outputs
In this example the output of the delegate is the output of quantized Conv. In XNNPACK, we must designate the output of Conv to be an external value. In this graph, however, the q node is the output. Since q is an implicit node (it does not actually running quantize) We pass its externalness to the conv node.
Inputs
i.e.
In this example q is the placeholder node and is given as input to the first dq. But we are running a quantized conv within the delegate. Since dq is implicit, we pass the externalness of the q place holder to dq, so that the input to Conv is external.
Reviewed By: digantdesai
Differential Revision: D48667676