Possible ways to debug XLA path in Tensorflow

[Aditya Atluri]

Hi,
I am seeking wisdom from developers who worked on XLA willing to share most useful ways to debug (especially emit markers) when running a code snippet using tensorflow.

I am inserting LOG(INFO)<<"MSG"; where I think the code will run. And changing VLOG_IS_ON level to 1 ((lvl) <= 1)

Thanks!

[Peter Hawkins]

LOG(INFO) or one of its relatives is the way to go for debug printouts. Which messages are printed is controlled by a couple of environment variables (TF_CPP_MIN_LOG_LEVEL and TF_CPP_MIN_VLOG_LEVEL), see:

https://github.com/tensorflow/tensorflow/blob/master/tensorflow/core/platform/default/logging.cc#L118

I think by default LOG(INFO) messages will print but VLOG() messages will not.

LOG(ERROR) messages will certainly print by default.

Also useful is LOG(FATAL), which prints a message and terminates the program. It's hard to miss that one!

Hope that helps,

Peter

--
You received this message because you are subscribed to the Google Groups "XLA development" group.
To unsubscribe from this group and stop receiving emails from it, send an email to xla-dev+u...@googlegroups.com.
To post to this group, send email to xla...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/xla-dev/77578387-7a8d-424a-a896-78722561a413%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Aditya Atluri]

Thanks! How about dumping the whole path traced by xla? Is there a way to dump that?