This repository was archived by the owner on Nov 16, 2023. It is now read-only.
src: allow operators to access graph in initialization #227
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
hanbitmyths
approved these changes
Sep 9, 2020
| * @param opsets a list of opsets that exported from the model | ||
| * @param graph the completely initialized graph | ||
| */ | ||
| resolve(node: Graph.Node, opsets: ReadonlyArray<OpSet>, graph: Graph): Operator; |
Collaborator
There was a problem hiding this comment.
NIT: not that different, but make graph as optional?
|
|
||
| export interface Operator { | ||
| initialize(attributes: Attribute): void; | ||
| initialize(attributes: Attribute, node: Graph.Node, graph: Graph): void; |
Collaborator
There was a problem hiding this comment.
Same here. node and graph as optional?
hanbitmyths
approved these changes
Sep 16, 2020
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
By accessing the graph ( already resolved and finalized ) in operator's initialization, it is possible to access the tensor value of the operator's input, if it's the graph's initializer or const value. This will enable implementation of upcoming operators that want to deal with constant inputs.