-
Notifications
You must be signed in to change notification settings - Fork 6.8k
Conversation
|
||
test = mx.symbol.BatchNorm(data, fix_gamma=False, use_global_stats=True, axis=chaxis) | ||
check_numeric_gradient(test, in_location, xmean_std, numeric_eps=1e-2, rtol=0.2, atol=0.01) | ||
check_numeric_gradient(test, in_location, xmean_std, numeric_eps=1e-2, rtol=0.2, atol=1e-4) |
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.
This looks like a decrease to me.
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.
Yeah it is an increase :)
Is this newly failing? That's a pretty bad tolerance. Are the c++ unit tests being run? Do they pass? |
What I mean is, did it just recently start failing or it's been failing periodically? I didn't see it on the list of failing tests before. What's the failure rate now? Are the cpp unit tests running? Are they passing? Batch norm operator is tested exhaustingly in the cpp unit tests. |
@cjolivier01 I have seen this failing since last week >75%, we are not running CPP test :( |
Was it failing before last week? Because if it wans t, it's not the fault
of the test. Not falling without 0.01 error is a problem.
…On Sun, Sep 17, 2017 at 10:30 PM Gautam Kumar ***@***.***> wrote:
@cjolivier01 <https://github.com/cjolivier01> I have seen this failing
since last week >75%, we are not running CPP test :(
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#7926 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AKts_U8ql35ePx7kMUzR7cavmHY5cqI1ks5sjgADgaJpZM4PaUjN>
.
|
@cjolivier01 I saw the failure since last two weeks, haven't noticed before. |
The test has been failing continuously here is another instance today |
Whops wrong test fix. |
@piiswrong @nswamy @cjolivier01