cls_report =               precision    recall  f1-score   support

         0.0     0.9595    0.7634    0.8503        93
         1.0     0.8021    0.9390    0.8652        82
         2.0     0.7534    0.6875    0.7190        80
         3.0     1.0000    1.0000    1.0000        57
         4.0     0.8116    0.8615    0.8358        65
         5.0     0.9620    0.9870    0.9744        77
         6.0     0.9780    1.0000    0.9889        89
         7.0     0.9481    1.0000    0.9733        73
         8.0     1.0000    1.0000    1.0000        13
         9.0     1.0000    1.0000    1.0000         9

    accuracy                         0.9028       638
   macro avg     0.9215    0.9239    0.9207       638
weighted avg     0.9049    0.9028    0.9012       638

eval_loss = 0.2609917640686035
macro_f1 = 0.9206824280993045
macro_p = 0.9214660895755212
macro_r = 0.9238516699010404
mcc = 0.8905145977958459