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
Debugger V2 not working. Invalid argument: DebugNumericSummaryV2Op requires tensor_id to be less than or equal to (2^53). Given tensor_id:26 #43608
Comments
@jaimeff |
@Saduf2019 you can run it by executing:
In my specific case, I have run it with the following command:
|
Same problem here. Searched all over for a solution and can't find one. Any help would be appreciated. |
@Saduf2019 Does the approach suggested by @jaimeff solve your problem? I'm not able to reproduce your issue with either the latest tf-nightly (2.4.0-dev20201007) or tf 2.3.1. I'm using the command
|
I get the following using that command:
This is the output of 'pip list' that is related to tensorflow:
So apparently, tensorflow package 2.3.1 installed with pip doesn't have DebugV2 support? |
@jaimeff This may be an operating system-specific issue. I see you are using Windows. I failed to reproduce the issue on Linux. Can you try running this Python script directly (instead of using |
Yes, that's exactly what I did when I saw I didn't have the file. So I copied it and ran it. The full output is in the first message (click here) (the one that I posted to open this issue). To summarize I'm getting the error message |
I got the same error. You have got a fix? |
Yes, but now you don't have debugging information. Am I right? The problem is that we cannot use Debugger V2 on Windows 10. The whole purpose of this ticket is to figure out how to make it work. Of course, if you disable it the problem is gone :-D |
…indows - A test in debug_v2_ops_test previously called `np.power(2, 53)` without specifying dtype. As a result, the output had the int32 dtype on Windows and caused overflowing. This apparently does not happen on Linux or Mac. - This CL fixes that by explicitly specifying `dtype=np.int64` in the call. - In debug_events_write.cc, check for whether the DebugEventsWriter instance is initialized and return early if not so. - This resolves a directory-not-empty test failure in debug_events_writer_test on Windows This is a step towards fixing #43608 PiperOrigin-RevId: 349569528 Change-Id: I8112f8faebe662542e80c03d5d95e8e089446fe8
I can still reproduce this bug on tf 2.4.1 on windows. |
I ran into the same issue on Windows 10 with tf 2.3.0 |
I played around with the parameters. It seems that the debugger runs with the defaults. i.e.
|
Same problem here. Has anyone found the solution? |
on Win10 TF 2.4.1 with the same error message |
Getting same issue in TF 2.5.0 in Windows 10, any work around? |
As I mention only the NO_TENSOR debug_mode works. However, it not really a workaround if you want to get details on your tensors. Other than that you could try to set up WSL or use a Linux distro and enjoy the journey of setting up another TensorFlow environment. |
Same issues TF 2.5.1 on Windows 10. |
Hi, I have just run into this issue with Tensorflow 2.9.1 and windows 10. |
@jainmilan Could you try to use calling tf.config.run_functions_eagerly(True) that will make all invocations of tf.function run eagerly instead of running as a traced graph function. This can be useful for debugging. Please refer to this doc and let us know if it helps? |
This issue has been automatically marked as stale because it has no recent activity. It will be closed if no further activity occurs. Thank you. |
Closing as stale. Please reopen if you'd like to work on this further. |
This issue is still happening. If I run functions eagerly, no graph data is available (as well as no graph executions to select). All I have is a massive list of python execution blocks on the timeline that are impossible to reasonably navigate and don't yield any sort of useful information aside from the stack trace and barebones output. |
Encountered the same issue (tensorflow 2.10.1 on Windows 10). |
same here with tensorflow 2.10.1 on Windows 10 |
System information
I cannot make the example work with Debugger V2.
By executing the example from the link above I get the following output:
I have also tried to build my own example with no success, same error:
DebugNumericSummaryV2Op requires tensor_id to be less than or equal to (2^53)
The text was updated successfully, but these errors were encountered: