[TOPI] add basic scheduling for conv2d_transpose on x86 #3491
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
yzhliu
reviewed
Jul 5, 2019
| def traverse(op): | ||
| """Traverse operators from computation graph""" | ||
| # inline all one-to-one-mapping operators except the last stage (output) | ||
| if tag.is_broadcast(op.tag) or tag.is_injective(op.tag): |
There was a problem hiding this comment.
is_injective implies is_broadcast
| def _declaration_conv2d_transpose(cfg, data, kernel, strides, padding, out_dtype): | ||
| return _declaration_conv2d_transpose_impl(cfg, data, kernel, strides, padding, out_dtype) | ||
|
|
||
| def _declaration_conv2d_transpose_impl(cfg, data, kernel, strides, padding, out_dtype): |
There was a problem hiding this comment.
can we put this function to nn/conv2d_transpose.py and have nn.conv2d_transpose_nchw and _declaration_conv2d_transpose share the implement?
There was a problem hiding this comment.
Makes sense. But I am assuming that cfg will be used in the future within this function. If we move it to nn/conv2d_transpose.py, I am not sure how to deal with it. What do you suggest? Thanks!
There was a problem hiding this comment.
Anyway, I modified as you suggested and put a TODO for now.
|
@yidawang do you folks happen to have an E2E example of Mask R-CNN with TVM? We'd be interested in this at FB for our object detector work. |
yzhliu
approved these changes
Jul 7, 2019
|
Thanks @yidawang |
wweic
pushed a commit
to wweic/tvm
that referenced
this pull request
Jul 11, 2019
* initialize cond 2d transpose scheduling on x86 * refine the scheduler a bit * fix for lint * address review comments; remove duplicate code * fix lint
wweic
pushed a commit
to wweic/tvm
that referenced
this pull request
Jul 11, 2019
* initialize cond 2d transpose scheduling on x86 * refine the scheduler a bit * fix for lint * address review comments; remove duplicate code * fix lint
wweic
pushed a commit
to neo-ai/tvm
that referenced
this pull request
Jul 11, 2019
* initialize cond 2d transpose scheduling on x86 * refine the scheduler a bit * fix for lint * address review comments; remove duplicate code * fix lint
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.
As the title. For workload of mask r-cnn, the #FLOPS improves from 0.9G to 90G on c5.18xlarge instance of Amazon EC2.
Finer tuning and autotvm remain the future work.
@yzhliu @kevinthesun