[TVMC] End to end benchmarking #10225
Closed
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
tmoreau89
requested review from
Huyuwei,
jwfromm,
kazum,
leandron,
mbrookhart,
siju-samuel and
srkreddy1238
as code owners
tmoreau89
force-pushed
the
tvmc_benchmark_e2e
branch
from
dd29865 to
4c7411d
Compare
tmoreau89
force-pushed
the
tvmc_benchmark_e2e
branch
from
4c7411d to
34d70de
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
End to end benchmarking was added to the VM and graph executor in order to faithfully represent execution time of a graph module in order to account for data transfer overheads. This can be particularly significant on discrete GPUs, where PCI-E transfers are important to account for in a typical model serving deployment.
This PR proposes to modify the default measurement done by TVMC to always benchmark end to end execution time from CPU local device context.
Another option is to expose in TVMC a flag that lets the user perform end to end benchmarking. I recommend however not benchmarking without data transfer overheads as it presents an overly optimistic outlook on TVM performance.