-
Notifications
You must be signed in to change notification settings - Fork 6.8k
[Doc] Start the tutorials for MKL-DNN backend #14202
Conversation
Thanks @TaoLv it will be very convenient for the end-user :) @ZhennanQin please create INT8 doc in this folder after this PR merge. |
Because lot of outside pages reference the MKLDNN_README. |
41b0a34
to
c7e9198
Compare
@mxnet-label-bot add[Doc, Example, pr-awaiting-review] |
@aaronmarkham Could you help to review the change? |
43dbdde
to
e7640da
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It's a good start to improve the MKLDNN documentation :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If MKL-DNN is on by default with the pip packages, then does it make sense to have the default build instructions for Windows/VS2017 to say that OpenBLAS is assumed? Why wouldn't the instructions be updated to bring some kind of conformity with what is suggested by what is provided by the pip package? To be specific, should these instructions for Windows be updated to suggest that MKL-DNN is used (at a minimum) if not also MKL?
@aaronmarkham Thank you for your kind suggestion. Currently, |
@aaronmarkham If yes, do you think we can do that in another PR? In this PR, I just try to move the document and make it visible on the web page, not intending to change any content in it. |
friendly ping @aaronmarkham |
Any comments? @aaronmarkham |
Sorry, I thought I already replied. We can do the docs update in a separate PR, but I wanted to flag that it should be done soon. I was under the impression that MKL-DNN was on by default. I guess this must be in the build from source route then? I ran into those subprocess issues. Should this be mentioned somewhere? Add a mention of Help me untangle what is what. |
Yes, MKL-DNN is on by default when build from source. For pypi package
I thought it's an issue of iomp not mxnet and someone is already working on it. I will track the progress of it. If it does not seem to be fixed soon, I will add it somewhere in the document.
MXNet should run into openblas and native CPU implementation. MKL and MKL-DNN code is not compiled into
You should explicitly use
MXNet has its bundled MKL-DNN sub-module. The sub-module will be compiled and linked when
Should share the similar methodology. One exception is that currently MKL-DNN is enabled by default only when build from source on linux + x86. |
@aaronmarkham I have created #14399 to track the document improvement effort. Do you think this PR is good to merge and we can move forward now? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM: ) Thank you for your contribution!
@aaronmarkham do you have any other suggestion for our plan? If this is no other comment, I will merge this PR in 24 hours as the first step of doc enhancement. |
I think the separate issue to track the docs needs is great! Go ahead with this one. |
Thanks, @aaronmarkham I am merging now :) |
Thank you @aaronmarkham . Feel free to comment in #14399 if you have any suggestion or question about MKL-DNN documentation. |
* start tutorials for mkldnn backend * fix links * add to tutorial index * fix sanity test * fix sanity test complain * add to webpage * Hint message and avoid 404
* start tutorials for mkldnn backend * fix links * add to tutorial index * fix sanity test * fix sanity test complain * add to webpage * Hint message and avoid 404
* start tutorials for mkldnn backend * fix links * add to tutorial index * fix sanity test * fix sanity test complain * add to webpage * Hint message and avoid 404
* start tutorials for mkldnn backend * fix links * add to tutorial index * fix sanity test * fix sanity test complain * add to webpage * Hint message and avoid 404
Description
This is the beginning of a serial of PRs for improving the tutorials of MKL-DNN backend, including installation, design, usage, quantization, performance faq, etc.
This PR moved the MKLDNN_README file from root folder to the new tutorial folder and fixed some links according to the change.
Checklist
Essentials
Please feel free to remove inapplicable items for your PR.
Changes
Comments