Different results with the current code

[orpheus92]

Hi, I have question with the current version of the code. I am running it on a google colab. However, for some reason I always got pretty low accuracy and comparatively high f-score. Do you know why that might happen? Those are just the results I got during training after the last epoch.
{'mtl_train(eval_set=val)': {'bert-base-uncased': {'InfoType': {'Accuracy': 0.2573,
'micro-F1': 0.6256,
'macro-F1': 0.4779,
'weighted-F1': 0.6176,
'macro-Precision': 0.5659,
'macro-Recall': 0.4356,
'micro-Precision': 0.6551,
'micro-Recall': 0.5986,
'weighted-Precision': 0.6459,
'weighted-Recall': 0.5986,
'support': 8433},
'Priority': {'Accuracy': 0.4664,
'micro-F1': 0.4664,
'macro-F1': 0.3757,
'weighted-F1': 0.5308,
'macro-Precision': 0.4096,
'macro-Recall': 0.4555,
'micro-Precision': 0.4664,
'micro-Recall': 0.4664,
'weighted-Precision': 0.7476,
'weighted-Recall': 0.4664,
'support': 3381}}}}