Auto-tuning a Convolutional Network for ARM CPU (tutorial error, bug reports) #8103
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
leandron
reviewed
May 24, 2021
There was a problem hiding this comment.
Hi @cbswj, thanks for the PR. I see you got some formatting issues. You can find some guidance on how to lint you patches here: https://tvm.apache.org/docs/contribute/pull_request.html#submit-a-pull-request
Regarding your question about the API and supported options for tuning, @jcf94 @comaniac can you help?
|
gentle ping to resolve the lint error, you can run |
|
@cbswj Hi, it nees you to fix the lint error following the I can help do that if the author didn't come back after several days. |
trevor-m
pushed a commit
to trevor-m/tvm
that referenced
this pull request
Jun 17, 2021
…reports) (apache#8103) * tune_relay_arm.py tutorial modify * Lint fix * Re-trigger CI Co-authored-by: Chenfan <[email protected]>
trevor-m
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Jun 17, 2021
…reports) (apache#8103) * tune_relay_arm.py tutorial modify * Lint fix * Re-trigger CI Co-authored-by: Chenfan <[email protected]>
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Outdated suggestions cannot be applied.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.
issue link : #8067
It is a problem when using RPC session to connect Raspberry Pi and GPU server to proceed autoTVM.
There was a problem with tuning on autoTVM.
As in the tutorial example, creating xgbtuner causes problems.
I looked up tvm-API to solve the problem and found this sentence.
‘For cross-device or cross-operator tuning, you can use ‘curve’ only.’
However, the tutorial creates an xgbtuner that does not have a curve option.
In addition, the tutorial doesn't even have code to create xgbtuner with curve options.
Although it is an autoTVM tutorial for ARM CPU using RPC session, it seems strange that there is no curve option when creating xgbtuner.
So I made all the feature_type options available by xgbtuner and experimented.
Of the 3 options knob, itervar, and curve, only knob option does not fail and works well!
Question
The API says to use the curve option when using xgb, but in reality, it is strange that the knob option, not the curve option, works!