Perform constant propagation when loading models #241
Merged
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
robertknight
force-pushed
the
const-prop-optimize
branch
2 times, most recently
from
cf31d1a to
4a7cfe5
Compare
This causes parts of the graph that depend only on constant inputs (ie. weights, biases) to be replaced when the model is loaded.
This optimization is now performed as part of graph optimizations inside `Model::load`, so does not need to be done by the CLI.
Since constant propagation is now performed as a graph optimization when the model is loaded, it only needs to be re-done if additional constants are added. This change assumes that graph optimizations were enabled when the model was loaded. If they were disabled, and no other constants were added, this could lead to expensive re-evaluation of unchanging parts of the graph on each run.
robertknight
force-pushed
the
const-prop-optimize
branch
from
4a7cfe5 to
75570a9
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Outdated suggestions cannot be applied.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.
Add initial infrastructure to perform graph optimizations as part of loading a model, and implement constant propagation (pre-evaluating the parts of the graph that don't depend on dynamic input values) as the first graph-level optimization. This is an optimization that was already implemented downstream in
rten-cliandrten-generateafter the model was loaded. Those downstream implementations are also removed in this change.Optimization is performed by default when a model is loaded, but can be disabled using
ModelOptions. This mirrors ONNX Runtime's approach of enabling optimizations unless the consumer opts out.TODO: