Add --n_iters flag to CLI and perform constant propagation before running model
#202
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
This is useful when running model inference in a loop, with a fixed set of options that have been passed as a parameter to the function running the loop.
This runs the model repeatedly. This is useful for assessing how much performance varies on the initial run vs after warmup is completed.
When testing `decoder_model.onnx` from https://huggingface.co/Mozilla/distilvit/tree/main/onnx I found that it was very slow in RTen compared to ONNX Runtime (>100ms vs 20ms). The main culprit turned out to be an expensive `Transpose` operator with constant inputs. ORT does constant propagation as part of its automatic optimizations when the model is loaded. RTen doesn't have an optimization like this implemented, but by implementing it "manually" in the CLI, we can see how much benefit this would provide.
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Outdated suggestions cannot be applied.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.
n_itersflag is useful for assessing the initial run vs "warmed up" inference times for models