Yes it shows on the UI and has the first epoch for some of the metrics but that's it. It has run like 50 epochs, it says it is still running but there are no updates to the scalars or debug samples

We are running the same code on multiple machines and it just randomly happens. Currently we are having the issue on 1 out of 4

I'll update my clearml version. Unfortunately I do not have a small code snippet and it is not always repeatable. Is there some additional logging that can be turned on?

Can you try with auto_connect_streams=True ? Also, what version of clearml sdk are you using?

Not sure if this is helpful but this is what I get when I cntrl-c out of the hung script

^C^CException ignored in atexit callback: <bound method Reporter._handle_program_exit of <clearml.backend_interface.metrics.reporter.Reporter object at 0x70fd8b7ff1c0>>
Event reporting sub-process lost, switching to thread based reporting
Traceback (most recent call last):
  File "/home/richard/.virtualenvs/temp_clearml/lib/python3.10/site-packages/clearml/backend_interface/metrics/reporter.py", line 317, in _handle_program_exit
    self.wait_for_events()
  File "/home/richard/.virtualenvs/temp_clearml/lib/python3.10/site-packages/clearml/backend_interface/metrics/reporter.py", line 337, in wait_for_events
    return report_service.wait_for_events(timeout=timeout)
  File "/home/richard/.virtualenvs/temp_clearml/lib/python3.10/site-packages/clearml/backend_interface/metrics/reporter.py", line 129, in wait_for_events
    if self._empty_state_event.wait(timeout=1.0):
  File "/home/richard/.virtualenvs/temp_clearml/lib/python3.10/site-packages/clearml/utilities/process/mp.py", line 445, in wait
    return self._event.wait(timeout=timeout)
  File "/usr/lib/python3.10/multiprocessing/synchronize.py", line 349, in wait
    self._cond.wait(timeout)
  File "/usr/lib/python3.10/multiprocessing/synchronize.py", line 261, in wait
    return self._wait_semaphore.acquire(True, timeout)
KeyboardInterrupt: 

My bad, if you set auto_connect_streams to false, you basically disable the console logging... Please see the documentation:

auto_connect_streams (Union[bool, Mapping[str, bool]]) – Control the automatic logging of stdout and stderr.

Can someone help with this?

Is this just the console output while training?

@<1719524641879363584:profile|ThankfulClams64> you could try using the compare function in the UI to compare the experiments on the machine the scalars are not reported properly and the experiments on a machine that runs the experiments properly. I suggest then replicating the environment exactly on the problematic machine. None

Correct, so I get something like this

ClearML Task: created new task id=6ec57dcb007545aebc4ec51eb5b34c67
======> WARNING! Git diff too large to store (2536kb), skipping uncommitted changes <======
ClearML results page: 

but that is all

@<1719524641879363584:profile|ThankfulClams64> , if you set auto_connect_streams to false nothing will be reported from your frameworks. With what frameworks are you working, tensorboard?

@<1719524641879363584:profile|ThankfulClams64> , can you provide a small code snippet that reproduces this behaviour? Can you also test with the latest version of clearml ?

I do have uncommitted code changes. I can try to check at some point if it would not have the problem without them. It seems like it could be repeated just by making a git repo with that script and adding a very large file. If I can repeat it is it best to open an issue in GitHub?

Hi @<1719524641879363584:profile|ThankfulClams64> , stopping all processes should do that, there is no programmatic way of doing that specifically. Did you try calling task.close() for all tasks you're using?

It seems similar to this None is it possible saving too many model weights causes metric logging thread to die?

Do you also see the same in the terminal itself on the machine?

Does any exit code appear? What is the status message and status reason in the 'INFO' section?

There is clearly some connection to the ClearML server as it remains "running" the entire training session but there are no metrics or debug samples. And I see nothing in the logs to indicate there is an issue

That makes sense... If you turn auto_connect_streams to false this mean that auto reporting will be disabled as per the documentation.. If you turn it to True then logging should resume.

The console logging still works. Aborting the task was in the log but did not work and the process continued until I killed it.

Yes tensorboard. It is still logging the tensorboard scalers and images. It just doesn't log the console output

The same training works sometimes. But I'm not sure how to troubleshoot when it stops logging the metrics

So I was able to repeat the same behavior on a machine running this example None

by adding the following callback

class TensorBoardImage(TensorBoard):
    @staticmethod
    def make_image(tensor):
        from PIL import Image
        import io
        tensor = np.stack((tensor, tensor, tensor), axis=2)
        height, width, channels = tensor.shape
        image = Image.fromarray(tensor)
        output = io.BytesIO()
        image.save(output, format='PNG')
        image_string = output.getvalue()
        output.close()
        return tf.Summary.Image(height=height,
                                width=width,
                                colorspace=channels,
                                encoded_image_string=image_string)

    def on_epoch_end(self, epoch, logs=None):
        if logs is None:
            logs = {}
        super(TensorBoardImage, self).on_epoch_end(epoch, logs)
        images = self.validation_data[0]  # 0 - data; 1 - labels
        img = (255 * images[0].reshape(28, 28)).astype('uint8')

        image = self.make_image(img)
        summary = tf.Summary(value=[tf.Summary.Value(tag='image', image=image)])
        self.writer.add_summary(summary, epoch)

So it seems like there is some bug in the how ClearML is logging tensorbaord images that causes everything to fail

It is still getting stuck. I think the issue might have something to do with the iterations versus epochs. I notice that one of the scalars that gets logged early is logging the epoch while the remaining scalars seem to be iterations because the iteration value is 1355 instead of 26

I am still having this issue. An update is that the "abort" does not work. Even though the state is correctly tracked in ClearML when I try to abort the experiment through the UI it says it does it but the experiment remains running on the computer.

Okay I will do another run to capture the console output. We currently set auto_connect_streams to False to reduce the number of API calls. So there isn't really anything in the ClearML task page console section

Is there someway to kill all connections of a machine to the ClearML server this does seem to be related to restarting a task / running a new task quickly after a task fails or is aborted

I am on 1.16.2

    task = Task.init(project_name=model_config['ClearML']['project_name'],
                     task_name=model_config['ClearML']['task_name'],
                     continue_last_task=False,
                     auto_connect_streams=True)

I am using 1.15.0. Yes I can try with auto_connect_streams set to True I believe I will still have the issue

Hi @<1719524641879363584:profile|ThankfulClams64> ,the logging is done by a separate process, I'm pretty sure it's not terminating all of the sudden. Did you manage to get a full log of such an experiment to share?