Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[TOPI] Fix bug in Winograd on CUDA #4260

Merged
merged 2 commits into from Nov 6, 2019

Conversation

comaniac
Copy link
Contributor

@comaniac comaniac commented Nov 5, 2019

Several topics [1, 2, 3] in the discuss mention that the conv2d failed to pass the shape checking in the runtime after the conv2d has been tuned by AutoTVM. This PR investigated the reason and resolved the issue. (thanks the help from @kevinthesun, @Laurawly and @vinx13 with the investigation).

Investigation
All errors happen at the same dimension: the output image height (arg2.shape[2]).
For example in [1], the workload has input (1, 80, 73, 73) with stride=1 and padding=0, so the output shape should be (1, 192, 71, 71). However, it encountered the following error:

TVMError: Check failed: ret == 0 (-1 vs. 0) : Assert fail: (73 == int32(arg2.shape[2])), Argument arg2.shape[2] has an unsatisfied constraint

It means that somehow the output shape is set to 73 instead of 71 during scheduling. After digging into the code, we found that Winograd schedule overrides the strides and padding to be 1 regardless the input workload.

Modification
Accordingly, this PR modified two parts. First, we get the stride and padding directly from the input workload and check if it is valid like previous. This change passed the isolated example that yyding provided in the TVM discuss.

Second, the Winograd unittest uses fallback config, but the Winograd schedule will fallback to the direct template if the config is fallback. It means the alter layer for Winograd schedule is never tested. This PR also forced the fallback to be False to enable the testing, which was suggested by @vinx13.

[1] https://discuss.tvm.ai/t/auto-tune-error-occurs-during-inference-when-using-auto-tuned-schedule
[2] https://discuss.tvm.ai/t/error-float16-for-cuda-with-autotvm
[3] https://discuss.tvm.ai/t/graphruntime-module-run-failed-when-created-with-logfile-from-autotvm-tuning

@kevinthesun
Copy link
Contributor

@cbalint13

@cbalint13
Copy link
Contributor

@comaniac ,

Looks good to me (looked at winograd part).

Copy link
Contributor

@Laurawly Laurawly left a comment

Choose a reason for hiding this comment

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

LGTM

@comaniac
Copy link
Contributor Author

comaniac commented Nov 6, 2019

@vinx13 could you help merge it? Thanks.

@vinx13
Copy link
Member

vinx13 commented Nov 6, 2019

@tqchen seems I don't have write permission

@Laurawly Laurawly merged commit 7211c27 into apache:master Nov 6, 2019
@Laurawly
Copy link
Contributor

Laurawly commented Nov 6, 2019

Thanks @comaniac @vinx13, this is now merged.

@comaniac comaniac deleted the fix_winograd_cuda branch November 6, 2019 23:03
@tqchen
Copy link
Member

tqchen commented Nov 6, 2019

@vinx13 you should have permission now, if now, please check if you have linked your github account. I will send you instructions

@vinx13
Copy link
Member

vinx13 commented Nov 7, 2019

@tqchen I still don't have permission. I'm already in the Apache github org, is there anything I'm missing?

zxy844288792 pushed a commit to neo-ai/tvm that referenced this pull request Nov 13, 2019
* fix winograd

* move get padding after kernel transform
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

6 participants