Registered artifacts not updated correctly between task runs

[materight]

Describe the bug

When continuing an existing task and trying to add rows to a DataFrame after calling register_artifact, sometimes the artifact is not updated correctly.

To reproduce

import pandas as pd
from clearml import Task

task = Task.init(project_name='Test', task_name='test', continue_last_task=True)
if 'df' not in task.artifacts:
    df = pd.DataFrame(columns=['a'])
else:
    df = task.artifacts['df'].get()
task.register_artifact('df', df)

for _ in range(50):
    df.loc[len(df)] = {'a': len(df)}

task.flush(wait_for_uploads=True)
print('df size:', len(df))

Results, when running the script for the:

• 1st time: df size: 50
• 2nd time: df size: 100
• 3rd time: df size: 100

Expected behaviour

The 3rd time should print df size: 150. It seems that the results of the second execution are not stored or returned correctly. Calling task.artifacts['df'].get() always returns only the first 50 rows. However, in the dashboard I can see 100 rows stored:

Environment

• Server type: app.clear.ml
• ClearML SDK Version: 1.10.4
• Python Version: 3.10.10
• OS: Linux

[materight]

It was an issue with my code, changing task.artifacts['df'].get() to task.artifacts['df'].get(force_download=True) solved it.