Disconnecting issue again and again

[Abhishek5077]

Describe the current behavior
A clear and concise explanation of what is currently happening.

Describe the expected behavior
A clear and concise explanation of what you expected to happen.

What web browser you are using
(Chrome, Firefox, Safari, etc.)

Additional context
Link to a minimal, public, self-contained notebook that reproduces this issue.

• Share the file using your GitHub account using File > Save a copy as a GitHub Gist.
• or Share Drive notebooks using the Share button then 'Get Shareable Link'.

[Abhishek5077]

The tool is disconnecting again and again... and even when I type negative prompts, negativity is showing in the picture

[docdnp]

Here some screenshot of the issue described.

Here the log when the runtime is disconnected:
colab.research.google.com-1735933611601.log

[EvanWiederspan]

Can you share the notebook you were using? Does this happen for all notebooks, or just the one?

[docdnp]

Can you share the notebook you were using? Does this happen for all notebooks, or just the one?

I can't share the notebook itself. But I can share the code of the cell that is running when this behavior occurs:

training_history = None

if os.path.exists(result_training_log):
  model = load_model_results(result_model_file, result_training_log)
else:
  original_stdout = sys.stdout
  with open(result_training_log, "w") as f:
      sys.stdout = Tee(original_stdout, f)
      training_history = model.fit(train_dataset,
                                  validation_data=test_dataset,
                                  epochs=epochs,
                                  steps_per_epoch=steps_per_epoch,
                                  callbacks=all_callbacks,
                                  verbose=1)
  sys.stdout = original_stdout

Actually it's a keras model which runs for several hours. I've encountered the problem each time this kind of cells are executed.

[cperry-goog]

I see a bunch of socketio errors in the log, assuming this is related to socketio we're working on a larger fix that will resolve.