Skip to content
This repository has been archived by the owner on Nov 17, 2023. It is now read-only.

[1.x] Update MXNet-TRT docs with the new optimize_for API #19385

Merged
merged 1 commit into from
Oct 23, 2020

Conversation

Kh4L
Copy link
Contributor

@Kh4L Kh4L commented Oct 20, 2020

Description

Update MXNet-TRT docs with the new optimize_for API.
Also update the numbers, run on a V100 system.

@mxnet-bot
Copy link

Hey @Kh4L , Thanks for submitting the PR
All tests are already queued to run once. If tests fail, you can trigger one or more tests again with the following commands:

  • To trigger all jobs: @mxnet-bot run ci [all]
  • To trigger specific jobs: @mxnet-bot run ci [job1, job2]

CI supported jobs: [sanity, windows-gpu, unix-cpu, clang, centos-gpu, windows-cpu, website, unix-gpu, miscellaneous, centos-cpu, edge]


Note:
Only following 3 categories can trigger CI :PR Author, MXNet Committer, Jenkins Admin.
All CI tests must pass before the PR can be merged.

@lanking520 lanking520 added pr-awaiting-testing PR is reviewed and waiting CI build and test pr-work-in-progress PR is still work in progress and removed pr-awaiting-testing PR is reviewed and waiting CI build and test labels Oct 20, 2020
@lanking520 lanking520 added pr-awaiting-testing PR is reviewed and waiting CI build and test and removed pr-work-in-progress PR is still work in progress labels Oct 21, 2020
@Kh4L Kh4L changed the title Update MXNet-TRT docs with the new optimize_for API [1.x] Update MXNet-TRT docs with the new optimize_for API Oct 21, 2020
@lanking520 lanking520 added pr-work-in-progress PR is still work in progress and removed pr-awaiting-testing PR is reviewed and waiting CI build and test labels Oct 21, 2020

We can give a simple speed up by turning on TensorRT FP16. This optimization comes almost as a freebie and doesn't need any other use effort than adding the optimize_for parameter precision.

```
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: use ```python

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Done

Copy link
Contributor

@samskalicky samskalicky left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the update to the docs!

@lanking520 lanking520 added pr-awaiting-testing PR is reviewed and waiting CI build and test pr-work-in-progress PR is still work in progress and removed pr-work-in-progress PR is still work in progress pr-awaiting-testing PR is reviewed and waiting CI build and test labels Oct 21, 2020
@lanking520 lanking520 added the pr-awaiting-review PR is waiting for code review label Oct 21, 2020
@samskalicky
Copy link
Contributor

Just a reminder, we're waiting for an update to the installation instructions specific to building with TRT before merging.

@Kh4L
Copy link
Contributor Author

Kh4L commented Oct 22, 2020

@samskalicky I am still waiting for updates from Wei, who is still trying to build it.

I would suggest to merge this PR and I will make another one with the build instructions

# Warmup
for i in range(0, 1000):
out = model(x)
out[0].wait_to_read()
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I'd be tempted to use mx.nd.waitall() for consistency with the samples above.

@KellenSunderland
Copy link
Contributor

Thanks for the update on this Serge. Looks good to me. Small nit with code samples, but let's call it non-blocking.

@samskalicky samskalicky merged commit 360542a into apache:v1.x Oct 23, 2020
josephevans pushed a commit to josephevans/mxnet that referenced this pull request Dec 8, 2020
@Kh4L Kh4L mentioned this pull request Mar 5, 2021
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
pr-awaiting-review PR is waiting for code review
Projects
None yet
Development

Successfully merging this pull request may close these issues.

5 participants